1 <!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01//EN"
2 "http://www.w3.org/TR/html4/strict.dtd">
5 <title>LLVM Programmer's Manual</title>
6 <link rel="stylesheet" href="llvm.css" type="text/css">
10 <div class="doc_title">
11 LLVM Programmer's Manual
15 <li><a href="#introduction">Introduction</a></li>
16 <li><a href="#general">General Information</a>
18 <li><a href="#stl">The C++ Standard Template Library</a></li>
20 <li>The <tt>-time-passes</tt> option</li>
21 <li>How to use the LLVM Makefile system</li>
22 <li>How to write a regression test</li>
27 <li><a href="#apis">Important and useful LLVM APIs</a>
29 <li><a href="#isa">The <tt>isa<></tt>, <tt>cast<></tt>
30 and <tt>dyn_cast<></tt> templates</a> </li>
31 <li><a href="#DEBUG">The <tt>DEBUG()</tt> macro and <tt>-debug</tt>
34 <li><a href="#DEBUG_TYPE">Fine grained debug info with <tt>DEBUG_TYPE</tt>
35 and the <tt>-debug-only</tt> option</a> </li>
38 <li><a href="#Statistic">The <tt>Statistic</tt> class & <tt>-stats</tt>
41 <li>The <tt>InstVisitor</tt> template
42 <li>The general graph API
44 <li><a href="#ViewGraph">Viewing graphs while debugging code</a></li>
47 <li><a href="#datastructure">Picking the Right Data Structure for a Task</a>
49 <li><a href="#ds_sequential">Sequential Containers (std::vector, std::list, etc)</a>
51 <li><a href="#dss_fixedarrays">Fixed Size Arrays</a></li>
52 <li><a href="#dss_heaparrays">Heap Allocated Arrays</a></li>
53 <li><a href="#dss_smallvector">"llvm/ADT/SmallVector.h"</a></li>
54 <li><a href="#dss_vector"><vector></a></li>
55 <li><a href="#dss_deque"><deque></a></li>
56 <li><a href="#dss_list"><list></a></li>
57 <li><a href="#dss_ilist">llvm/ADT/ilist</a></li>
58 <li><a href="#dss_other">Other Sequential Container Options</a></li>
60 <li><a href="#ds_set">Set-Like Containers (std::set, SmallSet, SetVector, etc)</a>
62 <li><a href="#dss_sortedvectorset">A sorted 'vector'</a></li>
63 <li><a href="#dss_smallset">"llvm/ADT/SmallSet.h"</a></li>
64 <li><a href="#dss_smallptrset">"llvm/ADT/SmallPtrSet.h"</a></li>
65 <li><a href="#dss_denseset">"llvm/ADT/DenseSet.h"</a></li>
66 <li><a href="#dss_FoldingSet">"llvm/ADT/FoldingSet.h"</a></li>
67 <li><a href="#dss_set"><set></a></li>
68 <li><a href="#dss_setvector">"llvm/ADT/SetVector.h"</a></li>
69 <li><a href="#dss_uniquevector">"llvm/ADT/UniqueVector.h"</a></li>
70 <li><a href="#dss_otherset">Other Set-Like ContainerOptions</a></li>
72 <li><a href="#ds_map">Map-Like Containers (std::map, DenseMap, etc)</a>
74 <li><a href="#dss_sortedvectormap">A sorted 'vector'</a></li>
75 <li><a href="#dss_stringmap">"llvm/ADT/StringMap.h"</a></li>
76 <li><a href="#dss_indexedmap">"llvm/ADT/IndexedMap.h"</a></li>
77 <li><a href="#dss_densemap">"llvm/ADT/DenseMap.h"</a></li>
78 <li><a href="#dss_map"><map></a></li>
79 <li><a href="#dss_othermap">Other Map-Like Container Options</a></li>
81 <li><a href="#ds_bit">BitVector-like containers</a>
83 <li><a href="#dss_bitvector">A dense bitvector</a></li>
84 <li><a href="#dss_sparsebitvector">A sparse bitvector</a></li>
88 <li><a href="#common">Helpful Hints for Common Operations</a>
90 <li><a href="#inspection">Basic Inspection and Traversal Routines</a>
92 <li><a href="#iterate_function">Iterating over the <tt>BasicBlock</tt>s
93 in a <tt>Function</tt></a> </li>
94 <li><a href="#iterate_basicblock">Iterating over the <tt>Instruction</tt>s
95 in a <tt>BasicBlock</tt></a> </li>
96 <li><a href="#iterate_institer">Iterating over the <tt>Instruction</tt>s
97 in a <tt>Function</tt></a> </li>
98 <li><a href="#iterate_convert">Turning an iterator into a
99 class pointer</a> </li>
100 <li><a href="#iterate_complex">Finding call sites: a more
101 complex example</a> </li>
102 <li><a href="#calls_and_invokes">Treating calls and invokes
103 the same way</a> </li>
104 <li><a href="#iterate_chains">Iterating over def-use &
105 use-def chains</a> </li>
106 <li><a href="#iterate_preds">Iterating over predecessors &
107 successors of blocks</a></li>
110 <li><a href="#simplechanges">Making simple changes</a>
112 <li><a href="#schanges_creating">Creating and inserting new
113 <tt>Instruction</tt>s</a> </li>
114 <li><a href="#schanges_deleting">Deleting <tt>Instruction</tt>s</a> </li>
115 <li><a href="#schanges_replacing">Replacing an <tt>Instruction</tt>
116 with another <tt>Value</tt></a> </li>
117 <li><a href="#schanges_deletingGV">Deleting <tt>GlobalVariable</tt>s</a> </li>
121 <li>Working with the Control Flow Graph
123 <li>Accessing predecessors and successors of a <tt>BasicBlock</tt>
131 <li><a href="#advanced">Advanced Topics</a>
133 <li><a href="#TypeResolve">LLVM Type Resolution</a>
135 <li><a href="#BuildRecType">Basic Recursive Type Construction</a></li>
136 <li><a href="#refineAbstractTypeTo">The <tt>refineAbstractTypeTo</tt> method</a></li>
137 <li><a href="#PATypeHolder">The PATypeHolder Class</a></li>
138 <li><a href="#AbstractTypeUser">The AbstractTypeUser Class</a></li>
141 <li><a href="#SymbolTable">The <tt>ValueSymbolTable</tt> and <tt>TypeSymbolTable</tt> classes </a></li>
144 <li><a href="#coreclasses">The Core LLVM Class Hierarchy Reference</a>
146 <li><a href="#Type">The <tt>Type</tt> class</a> </li>
147 <li><a href="#Module">The <tt>Module</tt> class</a></li>
148 <li><a href="#Value">The <tt>Value</tt> class</a>
150 <li><a href="#User">The <tt>User</tt> class</a>
152 <li><a href="#Instruction">The <tt>Instruction</tt> class</a></li>
153 <li><a href="#Constant">The <tt>Constant</tt> class</a>
155 <li><a href="#GlobalValue">The <tt>GlobalValue</tt> class</a>
157 <li><a href="#Function">The <tt>Function</tt> class</a></li>
158 <li><a href="#GlobalVariable">The <tt>GlobalVariable</tt> class</a></li>
165 <li><a href="#BasicBlock">The <tt>BasicBlock</tt> class</a></li>
166 <li><a href="#Argument">The <tt>Argument</tt> class</a></li>
173 <div class="doc_author">
174 <p>Written by <a href="mailto:sabre@nondot.org">Chris Lattner</a>,
175 <a href="mailto:dhurjati@cs.uiuc.edu">Dinakar Dhurjati</a>,
176 <a href="mailto:jstanley@cs.uiuc.edu">Joel Stanley</a>, and
177 <a href="mailto:rspencer@x10sys.com">Reid Spencer</a></p>
180 <!-- *********************************************************************** -->
181 <div class="doc_section">
182 <a name="introduction">Introduction </a>
184 <!-- *********************************************************************** -->
186 <div class="doc_text">
188 <p>This document is meant to highlight some of the important classes and
189 interfaces available in the LLVM source-base. This manual is not
190 intended to explain what LLVM is, how it works, and what LLVM code looks
191 like. It assumes that you know the basics of LLVM and are interested
192 in writing transformations or otherwise analyzing or manipulating the
195 <p>This document should get you oriented so that you can find your
196 way in the continuously growing source code that makes up the LLVM
197 infrastructure. Note that this manual is not intended to serve as a
198 replacement for reading the source code, so if you think there should be
199 a method in one of these classes to do something, but it's not listed,
200 check the source. Links to the <a href="/doxygen/">doxygen</a> sources
201 are provided to make this as easy as possible.</p>
203 <p>The first section of this document describes general information that is
204 useful to know when working in the LLVM infrastructure, and the second describes
205 the Core LLVM classes. In the future this manual will be extended with
206 information describing how to use extension libraries, such as dominator
207 information, CFG traversal routines, and useful utilities like the <tt><a
208 href="/doxygen/InstVisitor_8h-source.html">InstVisitor</a></tt> template.</p>
212 <!-- *********************************************************************** -->
213 <div class="doc_section">
214 <a name="general">General Information</a>
216 <!-- *********************************************************************** -->
218 <div class="doc_text">
220 <p>This section contains general information that is useful if you are working
221 in the LLVM source-base, but that isn't specific to any particular API.</p>
225 <!-- ======================================================================= -->
226 <div class="doc_subsection">
227 <a name="stl">The C++ Standard Template Library</a>
230 <div class="doc_text">
232 <p>LLVM makes heavy use of the C++ Standard Template Library (STL),
233 perhaps much more than you are used to, or have seen before. Because of
234 this, you might want to do a little background reading in the
235 techniques used and capabilities of the library. There are many good
236 pages that discuss the STL, and several books on the subject that you
237 can get, so it will not be discussed in this document.</p>
239 <p>Here are some useful links:</p>
243 <li><a href="http://www.dinkumware.com/refxcpp.html">Dinkumware C++ Library
244 reference</a> - an excellent reference for the STL and other parts of the
245 standard C++ library.</li>
247 <li><a href="http://www.tempest-sw.com/cpp/">C++ In a Nutshell</a> - This is an
248 O'Reilly book in the making. It has a decent
250 Reference that rivals Dinkumware's, and is unfortunately no longer free since the book has been
253 <li><a href="http://www.parashift.com/c++-faq-lite/">C++ Frequently Asked
256 <li><a href="http://www.sgi.com/tech/stl/">SGI's STL Programmer's Guide</a> -
258 href="http://www.sgi.com/tech/stl/stl_introduction.html">Introduction to the
261 <li><a href="http://www.research.att.com/%7Ebs/C++.html">Bjarne Stroustrup's C++
264 <li><a href="http://64.78.49.204/">
265 Bruce Eckel's Thinking in C++, 2nd ed. Volume 2 Revision 4.0 (even better, get
270 <p>You are also encouraged to take a look at the <a
271 href="CodingStandards.html">LLVM Coding Standards</a> guide which focuses on how
272 to write maintainable code more than where to put your curly braces.</p>
276 <!-- ======================================================================= -->
277 <div class="doc_subsection">
278 <a name="stl">Other useful references</a>
281 <div class="doc_text">
284 <li><a href="http://www.psc.edu/%7Esemke/cvs_branches.html">CVS
285 Branch and Tag Primer</a></li>
286 <li><a href="http://www.fortran-2000.com/ArnaudRecipes/sharedlib.html">Using
287 static and shared libraries across platforms</a></li>
292 <!-- *********************************************************************** -->
293 <div class="doc_section">
294 <a name="apis">Important and useful LLVM APIs</a>
296 <!-- *********************************************************************** -->
298 <div class="doc_text">
300 <p>Here we highlight some LLVM APIs that are generally useful and good to
301 know about when writing transformations.</p>
305 <!-- ======================================================================= -->
306 <div class="doc_subsection">
307 <a name="isa">The <tt>isa<></tt>, <tt>cast<></tt> and
308 <tt>dyn_cast<></tt> templates</a>
311 <div class="doc_text">
313 <p>The LLVM source-base makes extensive use of a custom form of RTTI.
314 These templates have many similarities to the C++ <tt>dynamic_cast<></tt>
315 operator, but they don't have some drawbacks (primarily stemming from
316 the fact that <tt>dynamic_cast<></tt> only works on classes that
317 have a v-table). Because they are used so often, you must know what they
318 do and how they work. All of these templates are defined in the <a
319 href="/doxygen/Casting_8h-source.html"><tt>llvm/Support/Casting.h</tt></a>
320 file (note that you very rarely have to include this file directly).</p>
323 <dt><tt>isa<></tt>: </dt>
325 <dd><p>The <tt>isa<></tt> operator works exactly like the Java
326 "<tt>instanceof</tt>" operator. It returns true or false depending on whether
327 a reference or pointer points to an instance of the specified class. This can
328 be very useful for constraint checking of various sorts (example below).</p>
331 <dt><tt>cast<></tt>: </dt>
333 <dd><p>The <tt>cast<></tt> operator is a "checked cast" operation. It
334 converts a pointer or reference from a base class to a derived cast, causing
335 an assertion failure if it is not really an instance of the right type. This
336 should be used in cases where you have some information that makes you believe
337 that something is of the right type. An example of the <tt>isa<></tt>
338 and <tt>cast<></tt> template is:</p>
340 <div class="doc_code">
342 static bool isLoopInvariant(const <a href="#Value">Value</a> *V, const Loop *L) {
343 if (isa<<a href="#Constant">Constant</a>>(V) || isa<<a href="#Argument">Argument</a>>(V) || isa<<a href="#GlobalValue">GlobalValue</a>>(V))
346 // <i>Otherwise, it must be an instruction...</i>
347 return !L->contains(cast<<a href="#Instruction">Instruction</a>>(V)->getParent());
352 <p>Note that you should <b>not</b> use an <tt>isa<></tt> test followed
353 by a <tt>cast<></tt>, for that use the <tt>dyn_cast<></tt>
358 <dt><tt>dyn_cast<></tt>:</dt>
360 <dd><p>The <tt>dyn_cast<></tt> operator is a "checking cast" operation.
361 It checks to see if the operand is of the specified type, and if so, returns a
362 pointer to it (this operator does not work with references). If the operand is
363 not of the correct type, a null pointer is returned. Thus, this works very
364 much like the <tt>dynamic_cast<></tt> operator in C++, and should be
365 used in the same circumstances. Typically, the <tt>dyn_cast<></tt>
366 operator is used in an <tt>if</tt> statement or some other flow control
367 statement like this:</p>
369 <div class="doc_code">
371 if (<a href="#AllocationInst">AllocationInst</a> *AI = dyn_cast<<a href="#AllocationInst">AllocationInst</a>>(Val)) {
377 <p>This form of the <tt>if</tt> statement effectively combines together a call
378 to <tt>isa<></tt> and a call to <tt>cast<></tt> into one
379 statement, which is very convenient.</p>
381 <p>Note that the <tt>dyn_cast<></tt> operator, like C++'s
382 <tt>dynamic_cast<></tt> or Java's <tt>instanceof</tt> operator, can be
383 abused. In particular, you should not use big chained <tt>if/then/else</tt>
384 blocks to check for lots of different variants of classes. If you find
385 yourself wanting to do this, it is much cleaner and more efficient to use the
386 <tt>InstVisitor</tt> class to dispatch over the instruction type directly.</p>
390 <dt><tt>cast_or_null<></tt>: </dt>
392 <dd><p>The <tt>cast_or_null<></tt> operator works just like the
393 <tt>cast<></tt> operator, except that it allows for a null pointer as an
394 argument (which it then propagates). This can sometimes be useful, allowing
395 you to combine several null checks into one.</p></dd>
397 <dt><tt>dyn_cast_or_null<></tt>: </dt>
399 <dd><p>The <tt>dyn_cast_or_null<></tt> operator works just like the
400 <tt>dyn_cast<></tt> operator, except that it allows for a null pointer
401 as an argument (which it then propagates). This can sometimes be useful,
402 allowing you to combine several null checks into one.</p></dd>
406 <p>These five templates can be used with any classes, whether they have a
407 v-table or not. To add support for these templates, you simply need to add
408 <tt>classof</tt> static methods to the class you are interested casting
409 to. Describing this is currently outside the scope of this document, but there
410 are lots of examples in the LLVM source base.</p>
414 <!-- ======================================================================= -->
415 <div class="doc_subsection">
416 <a name="DEBUG">The <tt>DEBUG()</tt> macro and <tt>-debug</tt> option</a>
419 <div class="doc_text">
421 <p>Often when working on your pass you will put a bunch of debugging printouts
422 and other code into your pass. After you get it working, you want to remove
423 it, but you may need it again in the future (to work out new bugs that you run
426 <p> Naturally, because of this, you don't want to delete the debug printouts,
427 but you don't want them to always be noisy. A standard compromise is to comment
428 them out, allowing you to enable them if you need them in the future.</p>
430 <p>The "<tt><a href="/doxygen/Debug_8h-source.html">llvm/Support/Debug.h</a></tt>"
431 file provides a macro named <tt>DEBUG()</tt> that is a much nicer solution to
432 this problem. Basically, you can put arbitrary code into the argument of the
433 <tt>DEBUG</tt> macro, and it is only executed if '<tt>opt</tt>' (or any other
434 tool) is run with the '<tt>-debug</tt>' command line argument:</p>
436 <div class="doc_code">
438 DOUT << "I am here!\n";
442 <p>Then you can run your pass like this:</p>
444 <div class="doc_code">
446 $ opt < a.bc > /dev/null -mypass
447 <i><no output></i>
448 $ opt < a.bc > /dev/null -mypass -debug
453 <p>Using the <tt>DEBUG()</tt> macro instead of a home-brewed solution allows you
454 to not have to create "yet another" command line option for the debug output for
455 your pass. Note that <tt>DEBUG()</tt> macros are disabled for optimized builds,
456 so they do not cause a performance impact at all (for the same reason, they
457 should also not contain side-effects!).</p>
459 <p>One additional nice thing about the <tt>DEBUG()</tt> macro is that you can
460 enable or disable it directly in gdb. Just use "<tt>set DebugFlag=0</tt>" or
461 "<tt>set DebugFlag=1</tt>" from the gdb if the program is running. If the
462 program hasn't been started yet, you can always just run it with
467 <!-- _______________________________________________________________________ -->
468 <div class="doc_subsubsection">
469 <a name="DEBUG_TYPE">Fine grained debug info with <tt>DEBUG_TYPE</tt> and
470 the <tt>-debug-only</tt> option</a>
473 <div class="doc_text">
475 <p>Sometimes you may find yourself in a situation where enabling <tt>-debug</tt>
476 just turns on <b>too much</b> information (such as when working on the code
477 generator). If you want to enable debug information with more fine-grained
478 control, you define the <tt>DEBUG_TYPE</tt> macro and the <tt>-debug</tt> only
479 option as follows:</p>
481 <div class="doc_code">
483 DOUT << "No debug type\n";
485 #define DEBUG_TYPE "foo"
486 DOUT << "'foo' debug type\n";
488 #define DEBUG_TYPE "bar"
489 DOUT << "'bar' debug type\n";
491 #define DEBUG_TYPE ""
492 DOUT << "No debug type (2)\n";
496 <p>Then you can run your pass like this:</p>
498 <div class="doc_code">
500 $ opt < a.bc > /dev/null -mypass
501 <i><no output></i>
502 $ opt < a.bc > /dev/null -mypass -debug
507 $ opt < a.bc > /dev/null -mypass -debug-only=foo
509 $ opt < a.bc > /dev/null -mypass -debug-only=bar
514 <p>Of course, in practice, you should only set <tt>DEBUG_TYPE</tt> at the top of
515 a file, to specify the debug type for the entire module (if you do this before
516 you <tt>#include "llvm/Support/Debug.h"</tt>, you don't have to insert the ugly
517 <tt>#undef</tt>'s). Also, you should use names more meaningful than "foo" and
518 "bar", because there is no system in place to ensure that names do not
519 conflict. If two different modules use the same string, they will all be turned
520 on when the name is specified. This allows, for example, all debug information
521 for instruction scheduling to be enabled with <tt>-debug-type=InstrSched</tt>,
522 even if the source lives in multiple files.</p>
526 <!-- ======================================================================= -->
527 <div class="doc_subsection">
528 <a name="Statistic">The <tt>Statistic</tt> class & <tt>-stats</tt>
532 <div class="doc_text">
535 href="/doxygen/Statistic_8h-source.html">llvm/ADT/Statistic.h</a></tt>" file
536 provides a class named <tt>Statistic</tt> that is used as a unified way to
537 keep track of what the LLVM compiler is doing and how effective various
538 optimizations are. It is useful to see what optimizations are contributing to
539 making a particular program run faster.</p>
541 <p>Often you may run your pass on some big program, and you're interested to see
542 how many times it makes a certain transformation. Although you can do this with
543 hand inspection, or some ad-hoc method, this is a real pain and not very useful
544 for big programs. Using the <tt>Statistic</tt> class makes it very easy to
545 keep track of this information, and the calculated information is presented in a
546 uniform manner with the rest of the passes being executed.</p>
548 <p>There are many examples of <tt>Statistic</tt> uses, but the basics of using
549 it are as follows:</p>
552 <li><p>Define your statistic like this:</p>
554 <div class="doc_code">
556 #define <a href="#DEBUG_TYPE">DEBUG_TYPE</a> "mypassname" <i>// This goes before any #includes.</i>
557 STATISTIC(NumXForms, "The # of times I did stuff");
561 <p>The <tt>STATISTIC</tt> macro defines a static variable, whose name is
562 specified by the first argument. The pass name is taken from the DEBUG_TYPE
563 macro, and the description is taken from the second argument. The variable
564 defined ("NumXForms" in this case) acts like an unsigned integer.</p></li>
566 <li><p>Whenever you make a transformation, bump the counter:</p>
568 <div class="doc_code">
570 ++NumXForms; // <i>I did stuff!</i>
577 <p>That's all you have to do. To get '<tt>opt</tt>' to print out the
578 statistics gathered, use the '<tt>-stats</tt>' option:</p>
580 <div class="doc_code">
582 $ opt -stats -mypassname < program.bc > /dev/null
583 <i>... statistics output ...</i>
587 <p> When running <tt>opt</tt> on a C file from the SPEC benchmark
588 suite, it gives a report that looks like this:</p>
590 <div class="doc_code">
592 7646 bitcodewriter - Number of normal instructions
593 725 bitcodewriter - Number of oversized instructions
594 129996 bitcodewriter - Number of bitcode bytes written
595 2817 raise - Number of insts DCEd or constprop'd
596 3213 raise - Number of cast-of-self removed
597 5046 raise - Number of expression trees converted
598 75 raise - Number of other getelementptr's formed
599 138 raise - Number of load/store peepholes
600 42 deadtypeelim - Number of unused typenames removed from symtab
601 392 funcresolve - Number of varargs functions resolved
602 27 globaldce - Number of global variables removed
603 2 adce - Number of basic blocks removed
604 134 cee - Number of branches revectored
605 49 cee - Number of setcc instruction eliminated
606 532 gcse - Number of loads removed
607 2919 gcse - Number of instructions removed
608 86 indvars - Number of canonical indvars added
609 87 indvars - Number of aux indvars removed
610 25 instcombine - Number of dead inst eliminate
611 434 instcombine - Number of insts combined
612 248 licm - Number of load insts hoisted
613 1298 licm - Number of insts hoisted to a loop pre-header
614 3 licm - Number of insts hoisted to multiple loop preds (bad, no loop pre-header)
615 75 mem2reg - Number of alloca's promoted
616 1444 cfgsimplify - Number of blocks simplified
620 <p>Obviously, with so many optimizations, having a unified framework for this
621 stuff is very nice. Making your pass fit well into the framework makes it more
622 maintainable and useful.</p>
626 <!-- ======================================================================= -->
627 <div class="doc_subsection">
628 <a name="ViewGraph">Viewing graphs while debugging code</a>
631 <div class="doc_text">
633 <p>Several of the important data structures in LLVM are graphs: for example
634 CFGs made out of LLVM <a href="#BasicBlock">BasicBlock</a>s, CFGs made out of
635 LLVM <a href="CodeGenerator.html#machinebasicblock">MachineBasicBlock</a>s, and
636 <a href="CodeGenerator.html#selectiondag_intro">Instruction Selection
637 DAGs</a>. In many cases, while debugging various parts of the compiler, it is
638 nice to instantly visualize these graphs.</p>
640 <p>LLVM provides several callbacks that are available in a debug build to do
641 exactly that. If you call the <tt>Function::viewCFG()</tt> method, for example,
642 the current LLVM tool will pop up a window containing the CFG for the function
643 where each basic block is a node in the graph, and each node contains the
644 instructions in the block. Similarly, there also exists
645 <tt>Function::viewCFGOnly()</tt> (does not include the instructions), the
646 <tt>MachineFunction::viewCFG()</tt> and <tt>MachineFunction::viewCFGOnly()</tt>,
647 and the <tt>SelectionDAG::viewGraph()</tt> methods. Within GDB, for example,
648 you can usually use something like <tt>call DAG.viewGraph()</tt> to pop
649 up a window. Alternatively, you can sprinkle calls to these functions in your
650 code in places you want to debug.</p>
652 <p>Getting this to work requires a small amount of configuration. On Unix
653 systems with X11, install the <a href="http://www.graphviz.org">graphviz</a>
654 toolkit, and make sure 'dot' and 'gv' are in your path. If you are running on
655 Mac OS/X, download and install the Mac OS/X <a
656 href="http://www.pixelglow.com/graphviz/">Graphviz program</a>, and add
657 <tt>/Applications/Graphviz.app/Contents/MacOS/</tt> (or wherever you install
658 it) to your path. Once in your system and path are set up, rerun the LLVM
659 configure script and rebuild LLVM to enable this functionality.</p>
661 <p><tt>SelectionDAG</tt> has been extended to make it easier to locate
662 <i>interesting</i> nodes in large complex graphs. From gdb, if you
663 <tt>call DAG.setGraphColor(<i>node</i>, "<i>color</i>")</tt>, then the
664 next <tt>call DAG.viewGraph()</tt> would highlight the node in the
665 specified color (choices of colors can be found at <a
666 href="http://www.graphviz.org/doc/info/colors.html">colors</a>.) More
667 complex node attributes can be provided with <tt>call
668 DAG.setGraphAttrs(<i>node</i>, "<i>attributes</i>")</tt> (choices can be
669 found at <a href="http://www.graphviz.org/doc/info/attrs.html">Graph
670 Attributes</a>.) If you want to restart and clear all the current graph
671 attributes, then you can <tt>call DAG.clearGraphAttrs()</tt>. </p>
675 <!-- *********************************************************************** -->
676 <div class="doc_section">
677 <a name="datastructure">Picking the Right Data Structure for a Task</a>
679 <!-- *********************************************************************** -->
681 <div class="doc_text">
683 <p>LLVM has a plethora of data structures in the <tt>llvm/ADT/</tt> directory,
684 and we commonly use STL data structures. This section describes the trade-offs
685 you should consider when you pick one.</p>
688 The first step is a choose your own adventure: do you want a sequential
689 container, a set-like container, or a map-like container? The most important
690 thing when choosing a container is the algorithmic properties of how you plan to
691 access the container. Based on that, you should use:</p>
694 <li>a <a href="#ds_map">map-like</a> container if you need efficient look-up
695 of an value based on another value. Map-like containers also support
696 efficient queries for containment (whether a key is in the map). Map-like
697 containers generally do not support efficient reverse mapping (values to
698 keys). If you need that, use two maps. Some map-like containers also
699 support efficient iteration through the keys in sorted order. Map-like
700 containers are the most expensive sort, only use them if you need one of
701 these capabilities.</li>
703 <li>a <a href="#ds_set">set-like</a> container if you need to put a bunch of
704 stuff into a container that automatically eliminates duplicates. Some
705 set-like containers support efficient iteration through the elements in
706 sorted order. Set-like containers are more expensive than sequential
710 <li>a <a href="#ds_sequential">sequential</a> container provides
711 the most efficient way to add elements and keeps track of the order they are
712 added to the collection. They permit duplicates and support efficient
713 iteration, but do not support efficient look-up based on a key.
716 <li>a <a href="#ds_bit">bit</a> container provides an efficient way to store and
717 perform set operations on sets of numeric id's, while automatically
718 eliminating duplicates. Bit containers require a maximum of 1 bit for each
719 identifier you want to store.
724 Once the proper category of container is determined, you can fine tune the
725 memory use, constant factors, and cache behaviors of access by intelligently
726 picking a member of the category. Note that constant factors and cache behavior
727 can be a big deal. If you have a vector that usually only contains a few
728 elements (but could contain many), for example, it's much better to use
729 <a href="#dss_smallvector">SmallVector</a> than <a href="#dss_vector">vector</a>
730 . Doing so avoids (relatively) expensive malloc/free calls, which dwarf the
731 cost of adding the elements to the container. </p>
735 <!-- ======================================================================= -->
736 <div class="doc_subsection">
737 <a name="ds_sequential">Sequential Containers (std::vector, std::list, etc)</a>
740 <div class="doc_text">
741 There are a variety of sequential containers available for you, based on your
742 needs. Pick the first in this section that will do what you want.
745 <!-- _______________________________________________________________________ -->
746 <div class="doc_subsubsection">
747 <a name="dss_fixedarrays">Fixed Size Arrays</a>
750 <div class="doc_text">
751 <p>Fixed size arrays are very simple and very fast. They are good if you know
752 exactly how many elements you have, or you have a (low) upper bound on how many
756 <!-- _______________________________________________________________________ -->
757 <div class="doc_subsubsection">
758 <a name="dss_heaparrays">Heap Allocated Arrays</a>
761 <div class="doc_text">
762 <p>Heap allocated arrays (new[] + delete[]) are also simple. They are good if
763 the number of elements is variable, if you know how many elements you will need
764 before the array is allocated, and if the array is usually large (if not,
765 consider a <a href="#dss_smallvector">SmallVector</a>). The cost of a heap
766 allocated array is the cost of the new/delete (aka malloc/free). Also note that
767 if you are allocating an array of a type with a constructor, the constructor and
768 destructors will be run for every element in the array (re-sizable vectors only
769 construct those elements actually used).</p>
772 <!-- _______________________________________________________________________ -->
773 <div class="doc_subsubsection">
774 <a name="dss_smallvector">"llvm/ADT/SmallVector.h"</a>
777 <div class="doc_text">
778 <p><tt>SmallVector<Type, N></tt> is a simple class that looks and smells
779 just like <tt>vector<Type></tt>:
780 it supports efficient iteration, lays out elements in memory order (so you can
781 do pointer arithmetic between elements), supports efficient push_back/pop_back
782 operations, supports efficient random access to its elements, etc.</p>
784 <p>The advantage of SmallVector is that it allocates space for
785 some number of elements (N) <b>in the object itself</b>. Because of this, if
786 the SmallVector is dynamically smaller than N, no malloc is performed. This can
787 be a big win in cases where the malloc/free call is far more expensive than the
788 code that fiddles around with the elements.</p>
790 <p>This is good for vectors that are "usually small" (e.g. the number of
791 predecessors/successors of a block is usually less than 8). On the other hand,
792 this makes the size of the SmallVector itself large, so you don't want to
793 allocate lots of them (doing so will waste a lot of space). As such,
794 SmallVectors are most useful when on the stack.</p>
796 <p>SmallVector also provides a nice portable and efficient replacement for
801 <!-- _______________________________________________________________________ -->
802 <div class="doc_subsubsection">
803 <a name="dss_vector"><vector></a>
806 <div class="doc_text">
808 std::vector is well loved and respected. It is useful when SmallVector isn't:
809 when the size of the vector is often large (thus the small optimization will
810 rarely be a benefit) or if you will be allocating many instances of the vector
811 itself (which would waste space for elements that aren't in the container).
812 vector is also useful when interfacing with code that expects vectors :).
815 <p>One worthwhile note about std::vector: avoid code like this:</p>
817 <div class="doc_code">
820 std::vector<foo> V;
826 <p>Instead, write this as:</p>
828 <div class="doc_code">
830 std::vector<foo> V;
838 <p>Doing so will save (at least) one heap allocation and free per iteration of
843 <!-- _______________________________________________________________________ -->
844 <div class="doc_subsubsection">
845 <a name="dss_deque"><deque></a>
848 <div class="doc_text">
849 <p>std::deque is, in some senses, a generalized version of std::vector. Like
850 std::vector, it provides constant time random access and other similar
851 properties, but it also provides efficient access to the front of the list. It
852 does not guarantee continuity of elements within memory.</p>
854 <p>In exchange for this extra flexibility, std::deque has significantly higher
855 constant factor costs than std::vector. If possible, use std::vector or
856 something cheaper.</p>
859 <!-- _______________________________________________________________________ -->
860 <div class="doc_subsubsection">
861 <a name="dss_list"><list></a>
864 <div class="doc_text">
865 <p>std::list is an extremely inefficient class that is rarely useful.
866 It performs a heap allocation for every element inserted into it, thus having an
867 extremely high constant factor, particularly for small data types. std::list
868 also only supports bidirectional iteration, not random access iteration.</p>
870 <p>In exchange for this high cost, std::list supports efficient access to both
871 ends of the list (like std::deque, but unlike std::vector or SmallVector). In
872 addition, the iterator invalidation characteristics of std::list are stronger
873 than that of a vector class: inserting or removing an element into the list does
874 not invalidate iterator or pointers to other elements in the list.</p>
877 <!-- _______________________________________________________________________ -->
878 <div class="doc_subsubsection">
879 <a name="dss_ilist">llvm/ADT/ilist</a>
882 <div class="doc_text">
883 <p><tt>ilist<T></tt> implements an 'intrusive' doubly-linked list. It is
884 intrusive, because it requires the element to store and provide access to the
885 prev/next pointers for the list.</p>
887 <p>ilist has the same drawbacks as std::list, and additionally requires an
888 ilist_traits implementation for the element type, but it provides some novel
889 characteristics. In particular, it can efficiently store polymorphic objects,
890 the traits class is informed when an element is inserted or removed from the
891 list, and ilists are guaranteed to support a constant-time splice operation.
894 <p>These properties are exactly what we want for things like Instructions and
895 basic blocks, which is why these are implemented with ilists.</p>
898 <!-- _______________________________________________________________________ -->
899 <div class="doc_subsubsection">
900 <a name="dss_other">Other Sequential Container options</a>
903 <div class="doc_text">
904 <p>Other STL containers are available, such as std::string.</p>
906 <p>There are also various STL adapter classes such as std::queue,
907 std::priority_queue, std::stack, etc. These provide simplified access to an
908 underlying container but don't affect the cost of the container itself.</p>
913 <!-- ======================================================================= -->
914 <div class="doc_subsection">
915 <a name="ds_set">Set-Like Containers (std::set, SmallSet, SetVector, etc)</a>
918 <div class="doc_text">
920 <p>Set-like containers are useful when you need to canonicalize multiple values
921 into a single representation. There are several different choices for how to do
922 this, providing various trade-offs.</p>
927 <!-- _______________________________________________________________________ -->
928 <div class="doc_subsubsection">
929 <a name="dss_sortedvectorset">A sorted 'vector'</a>
932 <div class="doc_text">
934 <p>If you intend to insert a lot of elements, then do a lot of queries, a
935 great approach is to use a vector (or other sequential container) with
936 std::sort+std::unique to remove duplicates. This approach works really well if
937 your usage pattern has these two distinct phases (insert then query), and can be
938 coupled with a good choice of <a href="#ds_sequential">sequential container</a>.
942 This combination provides the several nice properties: the result data is
943 contiguous in memory (good for cache locality), has few allocations, is easy to
944 address (iterators in the final vector are just indices or pointers), and can be
945 efficiently queried with a standard binary or radix search.</p>
949 <!-- _______________________________________________________________________ -->
950 <div class="doc_subsubsection">
951 <a name="dss_smallset">"llvm/ADT/SmallSet.h"</a>
954 <div class="doc_text">
956 <p>If you have a set-like data structure that is usually small and whose elements
957 are reasonably small, a <tt>SmallSet<Type, N></tt> is a good choice. This set
958 has space for N elements in place (thus, if the set is dynamically smaller than
959 N, no malloc traffic is required) and accesses them with a simple linear search.
960 When the set grows beyond 'N' elements, it allocates a more expensive representation that
961 guarantees efficient access (for most types, it falls back to std::set, but for
962 pointers it uses something far better, <a
963 href="#dss_smallptrset">SmallPtrSet</a>).</p>
965 <p>The magic of this class is that it handles small sets extremely efficiently,
966 but gracefully handles extremely large sets without loss of efficiency. The
967 drawback is that the interface is quite small: it supports insertion, queries
968 and erasing, but does not support iteration.</p>
972 <!-- _______________________________________________________________________ -->
973 <div class="doc_subsubsection">
974 <a name="dss_smallptrset">"llvm/ADT/SmallPtrSet.h"</a>
977 <div class="doc_text">
979 <p>SmallPtrSet has all the advantages of SmallSet (and a SmallSet of pointers is
980 transparently implemented with a SmallPtrSet), but also supports iterators. If
981 more than 'N' insertions are performed, a single quadratically
982 probed hash table is allocated and grows as needed, providing extremely
983 efficient access (constant time insertion/deleting/queries with low constant
984 factors) and is very stingy with malloc traffic.</p>
986 <p>Note that, unlike std::set, the iterators of SmallPtrSet are invalidated
987 whenever an insertion occurs. Also, the values visited by the iterators are not
988 visited in sorted order.</p>
992 <!-- _______________________________________________________________________ -->
993 <div class="doc_subsubsection">
994 <a name="dss_denseset">"llvm/ADT/DenseSet.h"</a>
997 <div class="doc_text">
1000 DenseSet is a simple quadratically probed hash table. It excels at supporting
1001 small values: it uses a single allocation to hold all of the pairs that
1002 are currently inserted in the set. DenseSet is a great way to unique small
1003 values that are not simple pointers (use <a
1004 href="#dss_smallptrset">SmallPtrSet</a> for pointers). Note that DenseSet has
1005 the same requirements for the value type that <a
1006 href="#dss_densemap">DenseMap</a> has.
1011 <!-- _______________________________________________________________________ -->
1012 <div class="doc_subsubsection">
1013 <a name="dss_FoldingSet">"llvm/ADT/FoldingSet.h"</a>
1016 <div class="doc_text">
1019 FoldingSet is an aggregate class that is really good at uniquing
1020 expensive-to-create or polymorphic objects. It is a combination of a chained
1021 hash table with intrusive links (uniqued objects are required to inherit from
1022 FoldingSetNode) that uses <a href="#dss_smallvector">SmallVector</a> as part of
1025 <p>Consider a case where you want to implement a "getOrCreateFoo" method for
1026 a complex object (for example, a node in the code generator). The client has a
1027 description of *what* it wants to generate (it knows the opcode and all the
1028 operands), but we don't want to 'new' a node, then try inserting it into a set
1029 only to find out it already exists, at which point we would have to delete it
1030 and return the node that already exists.
1033 <p>To support this style of client, FoldingSet perform a query with a
1034 FoldingSetNodeID (which wraps SmallVector) that can be used to describe the
1035 element that we want to query for. The query either returns the element
1036 matching the ID or it returns an opaque ID that indicates where insertion should
1037 take place. Construction of the ID usually does not require heap traffic.</p>
1039 <p>Because FoldingSet uses intrusive links, it can support polymorphic objects
1040 in the set (for example, you can have SDNode instances mixed with LoadSDNodes).
1041 Because the elements are individually allocated, pointers to the elements are
1042 stable: inserting or removing elements does not invalidate any pointers to other
1048 <!-- _______________________________________________________________________ -->
1049 <div class="doc_subsubsection">
1050 <a name="dss_set"><set></a>
1053 <div class="doc_text">
1055 <p><tt>std::set</tt> is a reasonable all-around set class, which is decent at
1056 many things but great at nothing. std::set allocates memory for each element
1057 inserted (thus it is very malloc intensive) and typically stores three pointers
1058 per element in the set (thus adding a large amount of per-element space
1059 overhead). It offers guaranteed log(n) performance, which is not particularly
1060 fast from a complexity standpoint (particularly if the elements of the set are
1061 expensive to compare, like strings), and has extremely high constant factors for
1062 lookup, insertion and removal.</p>
1064 <p>The advantages of std::set are that its iterators are stable (deleting or
1065 inserting an element from the set does not affect iterators or pointers to other
1066 elements) and that iteration over the set is guaranteed to be in sorted order.
1067 If the elements in the set are large, then the relative overhead of the pointers
1068 and malloc traffic is not a big deal, but if the elements of the set are small,
1069 std::set is almost never a good choice.</p>
1073 <!-- _______________________________________________________________________ -->
1074 <div class="doc_subsubsection">
1075 <a name="dss_setvector">"llvm/ADT/SetVector.h"</a>
1078 <div class="doc_text">
1079 <p>LLVM's SetVector<Type> is an adapter class that combines your choice of
1080 a set-like container along with a <a href="#ds_sequential">Sequential
1081 Container</a>. The important property
1082 that this provides is efficient insertion with uniquing (duplicate elements are
1083 ignored) with iteration support. It implements this by inserting elements into
1084 both a set-like container and the sequential container, using the set-like
1085 container for uniquing and the sequential container for iteration.
1088 <p>The difference between SetVector and other sets is that the order of
1089 iteration is guaranteed to match the order of insertion into the SetVector.
1090 This property is really important for things like sets of pointers. Because
1091 pointer values are non-deterministic (e.g. vary across runs of the program on
1092 different machines), iterating over the pointers in the set will
1093 not be in a well-defined order.</p>
1096 The drawback of SetVector is that it requires twice as much space as a normal
1097 set and has the sum of constant factors from the set-like container and the
1098 sequential container that it uses. Use it *only* if you need to iterate over
1099 the elements in a deterministic order. SetVector is also expensive to delete
1100 elements out of (linear time), unless you use it's "pop_back" method, which is
1104 <p>SetVector is an adapter class that defaults to using std::vector and std::set
1105 for the underlying containers, so it is quite expensive. However,
1106 <tt>"llvm/ADT/SetVector.h"</tt> also provides a SmallSetVector class, which
1107 defaults to using a SmallVector and SmallSet of a specified size. If you use
1108 this, and if your sets are dynamically smaller than N, you will save a lot of
1113 <!-- _______________________________________________________________________ -->
1114 <div class="doc_subsubsection">
1115 <a name="dss_uniquevector">"llvm/ADT/UniqueVector.h"</a>
1118 <div class="doc_text">
1121 UniqueVector is similar to <a href="#dss_setvector">SetVector</a>, but it
1122 retains a unique ID for each element inserted into the set. It internally
1123 contains a map and a vector, and it assigns a unique ID for each value inserted
1126 <p>UniqueVector is very expensive: its cost is the sum of the cost of
1127 maintaining both the map and vector, it has high complexity, high constant
1128 factors, and produces a lot of malloc traffic. It should be avoided.</p>
1133 <!-- _______________________________________________________________________ -->
1134 <div class="doc_subsubsection">
1135 <a name="dss_otherset">Other Set-Like Container Options</a>
1138 <div class="doc_text">
1141 The STL provides several other options, such as std::multiset and the various
1142 "hash_set" like containers (whether from C++ TR1 or from the SGI library).</p>
1144 <p>std::multiset is useful if you're not interested in elimination of
1145 duplicates, but has all the drawbacks of std::set. A sorted vector (where you
1146 don't delete duplicate entries) or some other approach is almost always
1149 <p>The various hash_set implementations (exposed portably by
1150 "llvm/ADT/hash_set") is a simple chained hashtable. This algorithm is as malloc
1151 intensive as std::set (performing an allocation for each element inserted,
1152 thus having really high constant factors) but (usually) provides O(1)
1153 insertion/deletion of elements. This can be useful if your elements are large
1154 (thus making the constant-factor cost relatively low) or if comparisons are
1155 expensive. Element iteration does not visit elements in a useful order.</p>
1159 <!-- ======================================================================= -->
1160 <div class="doc_subsection">
1161 <a name="ds_map">Map-Like Containers (std::map, DenseMap, etc)</a>
1164 <div class="doc_text">
1165 Map-like containers are useful when you want to associate data to a key. As
1166 usual, there are a lot of different ways to do this. :)
1169 <!-- _______________________________________________________________________ -->
1170 <div class="doc_subsubsection">
1171 <a name="dss_sortedvectormap">A sorted 'vector'</a>
1174 <div class="doc_text">
1177 If your usage pattern follows a strict insert-then-query approach, you can
1178 trivially use the same approach as <a href="#dss_sortedvectorset">sorted vectors
1179 for set-like containers</a>. The only difference is that your query function
1180 (which uses std::lower_bound to get efficient log(n) lookup) should only compare
1181 the key, not both the key and value. This yields the same advantages as sorted
1186 <!-- _______________________________________________________________________ -->
1187 <div class="doc_subsubsection">
1188 <a name="dss_stringmap">"llvm/ADT/StringMap.h"</a>
1191 <div class="doc_text">
1194 Strings are commonly used as keys in maps, and they are difficult to support
1195 efficiently: they are variable length, inefficient to hash and compare when
1196 long, expensive to copy, etc. StringMap is a specialized container designed to
1197 cope with these issues. It supports mapping an arbitrary range of bytes to an
1198 arbitrary other object.</p>
1200 <p>The StringMap implementation uses a quadratically-probed hash table, where
1201 the buckets store a pointer to the heap allocated entries (and some other
1202 stuff). The entries in the map must be heap allocated because the strings are
1203 variable length. The string data (key) and the element object (value) are
1204 stored in the same allocation with the string data immediately after the element
1205 object. This container guarantees the "<tt>(char*)(&Value+1)</tt>" points
1206 to the key string for a value.</p>
1208 <p>The StringMap is very fast for several reasons: quadratic probing is very
1209 cache efficient for lookups, the hash value of strings in buckets is not
1210 recomputed when lookup up an element, StringMap rarely has to touch the
1211 memory for unrelated objects when looking up a value (even when hash collisions
1212 happen), hash table growth does not recompute the hash values for strings
1213 already in the table, and each pair in the map is store in a single allocation
1214 (the string data is stored in the same allocation as the Value of a pair).</p>
1216 <p>StringMap also provides query methods that take byte ranges, so it only ever
1217 copies a string if a value is inserted into the table.</p>
1220 <!-- _______________________________________________________________________ -->
1221 <div class="doc_subsubsection">
1222 <a name="dss_indexedmap">"llvm/ADT/IndexedMap.h"</a>
1225 <div class="doc_text">
1227 IndexedMap is a specialized container for mapping small dense integers (or
1228 values that can be mapped to small dense integers) to some other type. It is
1229 internally implemented as a vector with a mapping function that maps the keys to
1230 the dense integer range.
1234 This is useful for cases like virtual registers in the LLVM code generator: they
1235 have a dense mapping that is offset by a compile-time constant (the first
1236 virtual register ID).</p>
1240 <!-- _______________________________________________________________________ -->
1241 <div class="doc_subsubsection">
1242 <a name="dss_densemap">"llvm/ADT/DenseMap.h"</a>
1245 <div class="doc_text">
1248 DenseMap is a simple quadratically probed hash table. It excels at supporting
1249 small keys and values: it uses a single allocation to hold all of the pairs that
1250 are currently inserted in the map. DenseMap is a great way to map pointers to
1251 pointers, or map other small types to each other.
1255 There are several aspects of DenseMap that you should be aware of, however. The
1256 iterators in a densemap are invalidated whenever an insertion occurs, unlike
1257 map. Also, because DenseMap allocates space for a large number of key/value
1258 pairs (it starts with 64 by default), it will waste a lot of space if your keys
1259 or values are large. Finally, you must implement a partial specialization of
1260 DenseMapInfo for the key that you want, if it isn't already supported. This
1261 is required to tell DenseMap about two special marker values (which can never be
1262 inserted into the map) that it needs internally.</p>
1266 <!-- _______________________________________________________________________ -->
1267 <div class="doc_subsubsection">
1268 <a name="dss_map"><map></a>
1271 <div class="doc_text">
1274 std::map has similar characteristics to <a href="#dss_set">std::set</a>: it uses
1275 a single allocation per pair inserted into the map, it offers log(n) lookup with
1276 an extremely large constant factor, imposes a space penalty of 3 pointers per
1277 pair in the map, etc.</p>
1279 <p>std::map is most useful when your keys or values are very large, if you need
1280 to iterate over the collection in sorted order, or if you need stable iterators
1281 into the map (i.e. they don't get invalidated if an insertion or deletion of
1282 another element takes place).</p>
1286 <!-- _______________________________________________________________________ -->
1287 <div class="doc_subsubsection">
1288 <a name="dss_othermap">Other Map-Like Container Options</a>
1291 <div class="doc_text">
1294 The STL provides several other options, such as std::multimap and the various
1295 "hash_map" like containers (whether from C++ TR1 or from the SGI library).</p>
1297 <p>std::multimap is useful if you want to map a key to multiple values, but has
1298 all the drawbacks of std::map. A sorted vector or some other approach is almost
1301 <p>The various hash_map implementations (exposed portably by
1302 "llvm/ADT/hash_map") are simple chained hash tables. This algorithm is as
1303 malloc intensive as std::map (performing an allocation for each element
1304 inserted, thus having really high constant factors) but (usually) provides O(1)
1305 insertion/deletion of elements. This can be useful if your elements are large
1306 (thus making the constant-factor cost relatively low) or if comparisons are
1307 expensive. Element iteration does not visit elements in a useful order.</p>
1311 <!-- ======================================================================= -->
1312 <div class="doc_subsection">
1313 <a name="ds_bit">Bit storage containers (BitVector, SparseBitVector)</a>
1316 <div class="doc_text">
1317 <p>Unlike the other containers, there are only two bit storage containers, and
1318 choosing when to use each is relatively straightforward.</p>
1320 <p>One additional option is
1321 <tt>std::vector<bool></tt>: we discourage its use for two reasons 1) the
1322 implementation in many common compilers (e.g. commonly available versions of
1323 GCC) is extremely inefficient and 2) the C++ standards committee is likely to
1324 deprecate this container and/or change it significantly somehow. In any case,
1325 please don't use it.</p>
1328 <!-- _______________________________________________________________________ -->
1329 <div class="doc_subsubsection">
1330 <a name="dss_bitvector">BitVector</a>
1333 <div class="doc_text">
1334 <p> The BitVector container provides a fixed size set of bits for manipulation.
1335 It supports individual bit setting/testing, as well as set operations. The set
1336 operations take time O(size of bitvector), but operations are performed one word
1337 at a time, instead of one bit at a time. This makes the BitVector very fast for
1338 set operations compared to other containers. Use the BitVector when you expect
1339 the number of set bits to be high (IE a dense set).
1343 <!-- _______________________________________________________________________ -->
1344 <div class="doc_subsubsection">
1345 <a name="dss_sparsebitvector">SparseBitVector</a>
1348 <div class="doc_text">
1349 <p> The SparseBitVector container is much like BitVector, with one major
1350 difference: Only the bits that are set, are stored. This makes the
1351 SparseBitVector much more space efficient than BitVector when the set is sparse,
1352 as well as making set operations O(number of set bits) instead of O(size of
1353 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
1354 (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).
1358 <!-- *********************************************************************** -->
1359 <div class="doc_section">
1360 <a name="common">Helpful Hints for Common Operations</a>
1362 <!-- *********************************************************************** -->
1364 <div class="doc_text">
1366 <p>This section describes how to perform some very simple transformations of
1367 LLVM code. This is meant to give examples of common idioms used, showing the
1368 practical side of LLVM transformations. <p> Because this is a "how-to" section,
1369 you should also read about the main classes that you will be working with. The
1370 <a href="#coreclasses">Core LLVM Class Hierarchy Reference</a> contains details
1371 and descriptions of the main classes that you should know about.</p>
1375 <!-- NOTE: this section should be heavy on example code -->
1376 <!-- ======================================================================= -->
1377 <div class="doc_subsection">
1378 <a name="inspection">Basic Inspection and Traversal Routines</a>
1381 <div class="doc_text">
1383 <p>The LLVM compiler infrastructure have many different data structures that may
1384 be traversed. Following the example of the C++ standard template library, the
1385 techniques used to traverse these various data structures are all basically the
1386 same. For a enumerable sequence of values, the <tt>XXXbegin()</tt> function (or
1387 method) returns an iterator to the start of the sequence, the <tt>XXXend()</tt>
1388 function returns an iterator pointing to one past the last valid element of the
1389 sequence, and there is some <tt>XXXiterator</tt> data type that is common
1390 between the two operations.</p>
1392 <p>Because the pattern for iteration is common across many different aspects of
1393 the program representation, the standard template library algorithms may be used
1394 on them, and it is easier to remember how to iterate. First we show a few common
1395 examples of the data structures that need to be traversed. Other data
1396 structures are traversed in very similar ways.</p>
1400 <!-- _______________________________________________________________________ -->
1401 <div class="doc_subsubsection">
1402 <a name="iterate_function">Iterating over the </a><a
1403 href="#BasicBlock"><tt>BasicBlock</tt></a>s in a <a
1404 href="#Function"><tt>Function</tt></a>
1407 <div class="doc_text">
1409 <p>It's quite common to have a <tt>Function</tt> instance that you'd like to
1410 transform in some way; in particular, you'd like to manipulate its
1411 <tt>BasicBlock</tt>s. To facilitate this, you'll need to iterate over all of
1412 the <tt>BasicBlock</tt>s that constitute the <tt>Function</tt>. The following is
1413 an example that prints the name of a <tt>BasicBlock</tt> and the number of
1414 <tt>Instruction</tt>s it contains:</p>
1416 <div class="doc_code">
1418 // <i>func is a pointer to a Function instance</i>
1419 for (Function::iterator i = func->begin(), e = func->end(); i != e; ++i)
1420 // <i>Print out the name of the basic block if it has one, and then the</i>
1421 // <i>number of instructions that it contains</i>
1422 llvm::cerr << "Basic block (name=" << i->getName() << ") has "
1423 << i->size() << " instructions.\n";
1427 <p>Note that i can be used as if it were a pointer for the purposes of
1428 invoking member functions of the <tt>Instruction</tt> class. This is
1429 because the indirection operator is overloaded for the iterator
1430 classes. In the above code, the expression <tt>i->size()</tt> is
1431 exactly equivalent to <tt>(*i).size()</tt> just like you'd expect.</p>
1435 <!-- _______________________________________________________________________ -->
1436 <div class="doc_subsubsection">
1437 <a name="iterate_basicblock">Iterating over the </a><a
1438 href="#Instruction"><tt>Instruction</tt></a>s in a <a
1439 href="#BasicBlock"><tt>BasicBlock</tt></a>
1442 <div class="doc_text">
1444 <p>Just like when dealing with <tt>BasicBlock</tt>s in <tt>Function</tt>s, it's
1445 easy to iterate over the individual instructions that make up
1446 <tt>BasicBlock</tt>s. Here's a code snippet that prints out each instruction in
1447 a <tt>BasicBlock</tt>:</p>
1449 <div class="doc_code">
1451 // <i>blk is a pointer to a BasicBlock instance</i>
1452 for (BasicBlock::iterator i = blk->begin(), e = blk->end(); i != e; ++i)
1453 // <i>The next statement works since operator<<(ostream&,...)</i>
1454 // <i>is overloaded for Instruction&</i>
1455 llvm::cerr << *i << "\n";
1459 <p>However, this isn't really the best way to print out the contents of a
1460 <tt>BasicBlock</tt>! Since the ostream operators are overloaded for virtually
1461 anything you'll care about, you could have just invoked the print routine on the
1462 basic block itself: <tt>llvm::cerr << *blk << "\n";</tt>.</p>
1466 <!-- _______________________________________________________________________ -->
1467 <div class="doc_subsubsection">
1468 <a name="iterate_institer">Iterating over the </a><a
1469 href="#Instruction"><tt>Instruction</tt></a>s in a <a
1470 href="#Function"><tt>Function</tt></a>
1473 <div class="doc_text">
1475 <p>If you're finding that you commonly iterate over a <tt>Function</tt>'s
1476 <tt>BasicBlock</tt>s and then that <tt>BasicBlock</tt>'s <tt>Instruction</tt>s,
1477 <tt>InstIterator</tt> should be used instead. You'll need to include <a
1478 href="/doxygen/InstIterator_8h-source.html"><tt>llvm/Support/InstIterator.h</tt></a>,
1479 and then instantiate <tt>InstIterator</tt>s explicitly in your code. Here's a
1480 small example that shows how to dump all instructions in a function to the standard error stream:<p>
1482 <div class="doc_code">
1484 #include "<a href="/doxygen/InstIterator_8h-source.html">llvm/Support/InstIterator.h</a>"
1486 // <i>F is a pointer to a Function instance</i>
1487 for (inst_iterator i = inst_begin(F), e = inst_end(F); i != e; ++i)
1488 llvm::cerr << *i << "\n";
1492 <p>Easy, isn't it? You can also use <tt>InstIterator</tt>s to fill a
1493 work list with its initial contents. For example, if you wanted to
1494 initialize a work list to contain all instructions in a <tt>Function</tt>
1495 F, all you would need to do is something like:</p>
1497 <div class="doc_code">
1499 std::set<Instruction*> worklist;
1500 worklist.insert(inst_begin(F), inst_end(F));
1504 <p>The STL set <tt>worklist</tt> would now contain all instructions in the
1505 <tt>Function</tt> pointed to by F.</p>
1509 <!-- _______________________________________________________________________ -->
1510 <div class="doc_subsubsection">
1511 <a name="iterate_convert">Turning an iterator into a class pointer (and
1515 <div class="doc_text">
1517 <p>Sometimes, it'll be useful to grab a reference (or pointer) to a class
1518 instance when all you've got at hand is an iterator. Well, extracting
1519 a reference or a pointer from an iterator is very straight-forward.
1520 Assuming that <tt>i</tt> is a <tt>BasicBlock::iterator</tt> and <tt>j</tt>
1521 is a <tt>BasicBlock::const_iterator</tt>:</p>
1523 <div class="doc_code">
1525 Instruction& inst = *i; // <i>Grab reference to instruction reference</i>
1526 Instruction* pinst = &*i; // <i>Grab pointer to instruction reference</i>
1527 const Instruction& inst = *j;
1531 <p>However, the iterators you'll be working with in the LLVM framework are
1532 special: they will automatically convert to a ptr-to-instance type whenever they
1533 need to. Instead of dereferencing the iterator and then taking the address of
1534 the result, you can simply assign the iterator to the proper pointer type and
1535 you get the dereference and address-of operation as a result of the assignment
1536 (behind the scenes, this is a result of overloading casting mechanisms). Thus
1537 the last line of the last example,</p>
1539 <div class="doc_code">
1541 Instruction *pinst = &*i;
1545 <p>is semantically equivalent to</p>
1547 <div class="doc_code">
1549 Instruction *pinst = i;
1553 <p>It's also possible to turn a class pointer into the corresponding iterator,
1554 and this is a constant time operation (very efficient). The following code
1555 snippet illustrates use of the conversion constructors provided by LLVM
1556 iterators. By using these, you can explicitly grab the iterator of something
1557 without actually obtaining it via iteration over some structure:</p>
1559 <div class="doc_code">
1561 void printNextInstruction(Instruction* inst) {
1562 BasicBlock::iterator it(inst);
1563 ++it; // <i>After this line, it refers to the instruction after *inst</i>
1564 if (it != inst->getParent()->end()) llvm::cerr << *it << "\n";
1571 <!--_______________________________________________________________________-->
1572 <div class="doc_subsubsection">
1573 <a name="iterate_complex">Finding call sites: a slightly more complex
1577 <div class="doc_text">
1579 <p>Say that you're writing a FunctionPass and would like to count all the
1580 locations in the entire module (that is, across every <tt>Function</tt>) where a
1581 certain function (i.e., some <tt>Function</tt>*) is already in scope. As you'll
1582 learn later, you may want to use an <tt>InstVisitor</tt> to accomplish this in a
1583 much more straight-forward manner, but this example will allow us to explore how
1584 you'd do it if you didn't have <tt>InstVisitor</tt> around. In pseudo-code, this
1585 is what we want to do:</p>
1587 <div class="doc_code">
1589 initialize callCounter to zero
1590 for each Function f in the Module
1591 for each BasicBlock b in f
1592 for each Instruction i in b
1593 if (i is a CallInst and calls the given function)
1594 increment callCounter
1598 <p>And the actual code is (remember, because we're writing a
1599 <tt>FunctionPass</tt>, our <tt>FunctionPass</tt>-derived class simply has to
1600 override the <tt>runOnFunction</tt> method):</p>
1602 <div class="doc_code">
1604 Function* targetFunc = ...;
1606 class OurFunctionPass : public FunctionPass {
1608 OurFunctionPass(): callCounter(0) { }
1610 virtual runOnFunction(Function& F) {
1611 for (Function::iterator b = F.begin(), be = F.end(); b != be; ++b) {
1612 for (BasicBlock::iterator i = b->begin(); ie = b->end(); i != ie; ++i) {
1613 if (<a href="#CallInst">CallInst</a>* callInst = <a href="#isa">dyn_cast</a><<a
1614 href="#CallInst">CallInst</a>>(&*i)) {
1615 // <i>We know we've encountered a call instruction, so we</i>
1616 // <i>need to determine if it's a call to the</i>
1617 // <i>function pointed to by m_func or not.</i>
1618 if (callInst->getCalledFunction() == targetFunc)
1626 unsigned callCounter;
1633 <!--_______________________________________________________________________-->
1634 <div class="doc_subsubsection">
1635 <a name="calls_and_invokes">Treating calls and invokes the same way</a>
1638 <div class="doc_text">
1640 <p>You may have noticed that the previous example was a bit oversimplified in
1641 that it did not deal with call sites generated by 'invoke' instructions. In
1642 this, and in other situations, you may find that you want to treat
1643 <tt>CallInst</tt>s and <tt>InvokeInst</tt>s the same way, even though their
1644 most-specific common base class is <tt>Instruction</tt>, which includes lots of
1645 less closely-related things. For these cases, LLVM provides a handy wrapper
1647 href="http://llvm.org/doxygen/classllvm_1_1CallSite.html"><tt>CallSite</tt></a>.
1648 It is essentially a wrapper around an <tt>Instruction</tt> pointer, with some
1649 methods that provide functionality common to <tt>CallInst</tt>s and
1650 <tt>InvokeInst</tt>s.</p>
1652 <p>This class has "value semantics": it should be passed by value, not by
1653 reference and it should not be dynamically allocated or deallocated using
1654 <tt>operator new</tt> or <tt>operator delete</tt>. It is efficiently copyable,
1655 assignable and constructable, with costs equivalents to that of a bare pointer.
1656 If you look at its definition, it has only a single pointer member.</p>
1660 <!--_______________________________________________________________________-->
1661 <div class="doc_subsubsection">
1662 <a name="iterate_chains">Iterating over def-use & use-def chains</a>
1665 <div class="doc_text">
1667 <p>Frequently, we might have an instance of the <a
1668 href="/doxygen/classllvm_1_1Value.html">Value Class</a> and we want to
1669 determine which <tt>User</tt>s use the <tt>Value</tt>. The list of all
1670 <tt>User</tt>s of a particular <tt>Value</tt> is called a <i>def-use</i> chain.
1671 For example, let's say we have a <tt>Function*</tt> named <tt>F</tt> to a
1672 particular function <tt>foo</tt>. Finding all of the instructions that
1673 <i>use</i> <tt>foo</tt> is as simple as iterating over the <i>def-use</i> chain
1676 <div class="doc_code">
1680 for (Value::use_iterator i = F->use_begin(), e = F->use_end(); i != e; ++i)
1681 if (Instruction *Inst = dyn_cast<Instruction>(*i)) {
1682 llvm::cerr << "F is used in instruction:\n";
1683 llvm::cerr << *Inst << "\n";
1688 <p>Alternately, it's common to have an instance of the <a
1689 href="/doxygen/classllvm_1_1User.html">User Class</a> and need to know what
1690 <tt>Value</tt>s are used by it. The list of all <tt>Value</tt>s used by a
1691 <tt>User</tt> is known as a <i>use-def</i> chain. Instances of class
1692 <tt>Instruction</tt> are common <tt>User</tt>s, so we might want to iterate over
1693 all of the values that a particular instruction uses (that is, the operands of
1694 the particular <tt>Instruction</tt>):</p>
1696 <div class="doc_code">
1698 Instruction *pi = ...;
1700 for (User::op_iterator i = pi->op_begin(), e = pi->op_end(); i != e; ++i) {
1708 def-use chains ("finding all users of"): Value::use_begin/use_end
1709 use-def chains ("finding all values used"): User::op_begin/op_end [op=operand]
1714 <!--_______________________________________________________________________-->
1715 <div class="doc_subsubsection">
1716 <a name="iterate_preds">Iterating over predecessors &
1717 successors of blocks</a>
1720 <div class="doc_text">
1722 <p>Iterating over the predecessors and successors of a block is quite easy
1723 with the routines defined in <tt>"llvm/Support/CFG.h"</tt>. Just use code like
1724 this to iterate over all predecessors of BB:</p>
1726 <div class="doc_code">
1728 #include "llvm/Support/CFG.h"
1729 BasicBlock *BB = ...;
1731 for (pred_iterator PI = pred_begin(BB), E = pred_end(BB); PI != E; ++PI) {
1732 BasicBlock *Pred = *PI;
1738 <p>Similarly, to iterate over successors use
1739 succ_iterator/succ_begin/succ_end.</p>
1744 <!-- ======================================================================= -->
1745 <div class="doc_subsection">
1746 <a name="simplechanges">Making simple changes</a>
1749 <div class="doc_text">
1751 <p>There are some primitive transformation operations present in the LLVM
1752 infrastructure that are worth knowing about. When performing
1753 transformations, it's fairly common to manipulate the contents of basic
1754 blocks. This section describes some of the common methods for doing so
1755 and gives example code.</p>
1759 <!--_______________________________________________________________________-->
1760 <div class="doc_subsubsection">
1761 <a name="schanges_creating">Creating and inserting new
1762 <tt>Instruction</tt>s</a>
1765 <div class="doc_text">
1767 <p><i>Instantiating Instructions</i></p>
1769 <p>Creation of <tt>Instruction</tt>s is straight-forward: simply call the
1770 constructor for the kind of instruction to instantiate and provide the necessary
1771 parameters. For example, an <tt>AllocaInst</tt> only <i>requires</i> a
1772 (const-ptr-to) <tt>Type</tt>. Thus:</p>
1774 <div class="doc_code">
1776 AllocaInst* ai = new AllocaInst(Type::Int32Ty);
1780 <p>will create an <tt>AllocaInst</tt> instance that represents the allocation of
1781 one integer in the current stack frame, at run time. Each <tt>Instruction</tt>
1782 subclass is likely to have varying default parameters which change the semantics
1783 of the instruction, so refer to the <a
1784 href="/doxygen/classllvm_1_1Instruction.html">doxygen documentation for the subclass of
1785 Instruction</a> that you're interested in instantiating.</p>
1787 <p><i>Naming values</i></p>
1789 <p>It is very useful to name the values of instructions when you're able to, as
1790 this facilitates the debugging of your transformations. If you end up looking
1791 at generated LLVM machine code, you definitely want to have logical names
1792 associated with the results of instructions! By supplying a value for the
1793 <tt>Name</tt> (default) parameter of the <tt>Instruction</tt> constructor, you
1794 associate a logical name with the result of the instruction's execution at
1795 run time. For example, say that I'm writing a transformation that dynamically
1796 allocates space for an integer on the stack, and that integer is going to be
1797 used as some kind of index by some other code. To accomplish this, I place an
1798 <tt>AllocaInst</tt> at the first point in the first <tt>BasicBlock</tt> of some
1799 <tt>Function</tt>, and I'm intending to use it within the same
1800 <tt>Function</tt>. I might do:</p>
1802 <div class="doc_code">
1804 AllocaInst* pa = new AllocaInst(Type::Int32Ty, 0, "indexLoc");
1808 <p>where <tt>indexLoc</tt> is now the logical name of the instruction's
1809 execution value, which is a pointer to an integer on the run time stack.</p>
1811 <p><i>Inserting instructions</i></p>
1813 <p>There are essentially two ways to insert an <tt>Instruction</tt>
1814 into an existing sequence of instructions that form a <tt>BasicBlock</tt>:</p>
1817 <li>Insertion into an explicit instruction list
1819 <p>Given a <tt>BasicBlock* pb</tt>, an <tt>Instruction* pi</tt> within that
1820 <tt>BasicBlock</tt>, and a newly-created instruction we wish to insert
1821 before <tt>*pi</tt>, we do the following: </p>
1823 <div class="doc_code">
1825 BasicBlock *pb = ...;
1826 Instruction *pi = ...;
1827 Instruction *newInst = new Instruction(...);
1829 pb->getInstList().insert(pi, newInst); // <i>Inserts newInst before pi in pb</i>
1833 <p>Appending to the end of a <tt>BasicBlock</tt> is so common that
1834 the <tt>Instruction</tt> class and <tt>Instruction</tt>-derived
1835 classes provide constructors which take a pointer to a
1836 <tt>BasicBlock</tt> to be appended to. For example code that
1839 <div class="doc_code">
1841 BasicBlock *pb = ...;
1842 Instruction *newInst = new Instruction(...);
1844 pb->getInstList().push_back(newInst); // <i>Appends newInst to pb</i>
1850 <div class="doc_code">
1852 BasicBlock *pb = ...;
1853 Instruction *newInst = new Instruction(..., pb);
1857 <p>which is much cleaner, especially if you are creating
1858 long instruction streams.</p></li>
1860 <li>Insertion into an implicit instruction list
1862 <p><tt>Instruction</tt> instances that are already in <tt>BasicBlock</tt>s
1863 are implicitly associated with an existing instruction list: the instruction
1864 list of the enclosing basic block. Thus, we could have accomplished the same
1865 thing as the above code without being given a <tt>BasicBlock</tt> by doing:
1868 <div class="doc_code">
1870 Instruction *pi = ...;
1871 Instruction *newInst = new Instruction(...);
1873 pi->getParent()->getInstList().insert(pi, newInst);
1877 <p>In fact, this sequence of steps occurs so frequently that the
1878 <tt>Instruction</tt> class and <tt>Instruction</tt>-derived classes provide
1879 constructors which take (as a default parameter) a pointer to an
1880 <tt>Instruction</tt> which the newly-created <tt>Instruction</tt> should
1881 precede. That is, <tt>Instruction</tt> constructors are capable of
1882 inserting the newly-created instance into the <tt>BasicBlock</tt> of a
1883 provided instruction, immediately before that instruction. Using an
1884 <tt>Instruction</tt> constructor with a <tt>insertBefore</tt> (default)
1885 parameter, the above code becomes:</p>
1887 <div class="doc_code">
1889 Instruction* pi = ...;
1890 Instruction* newInst = new Instruction(..., pi);
1894 <p>which is much cleaner, especially if you're creating a lot of
1895 instructions and adding them to <tt>BasicBlock</tt>s.</p></li>
1900 <!--_______________________________________________________________________-->
1901 <div class="doc_subsubsection">
1902 <a name="schanges_deleting">Deleting <tt>Instruction</tt>s</a>
1905 <div class="doc_text">
1907 <p>Deleting an instruction from an existing sequence of instructions that form a
1908 <a href="#BasicBlock"><tt>BasicBlock</tt></a> is very straight-forward. First,
1909 you must have a pointer to the instruction that you wish to delete. Second, you
1910 need to obtain the pointer to that instruction's basic block. You use the
1911 pointer to the basic block to get its list of instructions and then use the
1912 erase function to remove your instruction. For example:</p>
1914 <div class="doc_code">
1916 <a href="#Instruction">Instruction</a> *I = .. ;
1917 I->eraseFromParent();
1923 <!--_______________________________________________________________________-->
1924 <div class="doc_subsubsection">
1925 <a name="schanges_replacing">Replacing an <tt>Instruction</tt> with another
1929 <div class="doc_text">
1931 <p><i>Replacing individual instructions</i></p>
1933 <p>Including "<a href="/doxygen/BasicBlockUtils_8h-source.html">llvm/Transforms/Utils/BasicBlockUtils.h</a>"
1934 permits use of two very useful replace functions: <tt>ReplaceInstWithValue</tt>
1935 and <tt>ReplaceInstWithInst</tt>.</p>
1937 <h4><a name="schanges_deleting">Deleting <tt>Instruction</tt>s</a></h4>
1940 <li><tt>ReplaceInstWithValue</tt>
1942 <p>This function replaces all uses (within a basic block) of a given
1943 instruction with a value, and then removes the original instruction. The
1944 following example illustrates the replacement of the result of a particular
1945 <tt>AllocaInst</tt> that allocates memory for a single integer with a null
1946 pointer to an integer.</p>
1948 <div class="doc_code">
1950 AllocaInst* instToReplace = ...;
1951 BasicBlock::iterator ii(instToReplace);
1953 ReplaceInstWithValue(instToReplace->getParent()->getInstList(), ii,
1954 Constant::getNullValue(PointerType::get(Type::Int32Ty)));
1957 <li><tt>ReplaceInstWithInst</tt>
1959 <p>This function replaces a particular instruction with another
1960 instruction. The following example illustrates the replacement of one
1961 <tt>AllocaInst</tt> with another.</p>
1963 <div class="doc_code">
1965 AllocaInst* instToReplace = ...;
1966 BasicBlock::iterator ii(instToReplace);
1968 ReplaceInstWithInst(instToReplace->getParent()->getInstList(), ii,
1969 new AllocaInst(Type::Int32Ty, 0, "ptrToReplacedInt"));
1973 <p><i>Replacing multiple uses of <tt>User</tt>s and <tt>Value</tt>s</i></p>
1975 <p>You can use <tt>Value::replaceAllUsesWith</tt> and
1976 <tt>User::replaceUsesOfWith</tt> to change more than one use at a time. See the
1977 doxygen documentation for the <a href="/doxygen/classllvm_1_1Value.html">Value Class</a>
1978 and <a href="/doxygen/classllvm_1_1User.html">User Class</a>, respectively, for more
1981 <!-- Value::replaceAllUsesWith User::replaceUsesOfWith Point out:
1982 include/llvm/Transforms/Utils/ especially BasicBlockUtils.h with:
1983 ReplaceInstWithValue, ReplaceInstWithInst -->
1987 <!--_______________________________________________________________________-->
1988 <div class="doc_subsubsection">
1989 <a name="schanges_deletingGV">Deleting <tt>GlobalVariable</tt>s</a>
1992 <div class="doc_text">
1994 <p>Deleting a global variable from a module is just as easy as deleting an
1995 Instruction. First, you must have a pointer to the global variable that you wish
1996 to delete. You use this pointer to erase it from its parent, the module.
1999 <div class="doc_code">
2001 <a href="#GlobalVariable">GlobalVariable</a> *GV = .. ;
2003 GV->eraseFromParent();
2009 <!-- *********************************************************************** -->
2010 <div class="doc_section">
2011 <a name="advanced">Advanced Topics</a>
2013 <!-- *********************************************************************** -->
2015 <div class="doc_text">
2017 This section describes some of the advanced or obscure API's that most clients
2018 do not need to be aware of. These API's tend manage the inner workings of the
2019 LLVM system, and only need to be accessed in unusual circumstances.
2023 <!-- ======================================================================= -->
2024 <div class="doc_subsection">
2025 <a name="TypeResolve">LLVM Type Resolution</a>
2028 <div class="doc_text">
2031 The LLVM type system has a very simple goal: allow clients to compare types for
2032 structural equality with a simple pointer comparison (aka a shallow compare).
2033 This goal makes clients much simpler and faster, and is used throughout the LLVM
2038 Unfortunately achieving this goal is not a simple matter. In particular,
2039 recursive types and late resolution of opaque types makes the situation very
2040 difficult to handle. Fortunately, for the most part, our implementation makes
2041 most clients able to be completely unaware of the nasty internal details. The
2042 primary case where clients are exposed to the inner workings of it are when
2043 building a recursive type. In addition to this case, the LLVM bitcode reader,
2044 assembly parser, and linker also have to be aware of the inner workings of this
2049 For our purposes below, we need three concepts. First, an "Opaque Type" is
2050 exactly as defined in the <a href="LangRef.html#t_opaque">language
2051 reference</a>. Second an "Abstract Type" is any type which includes an
2052 opaque type as part of its type graph (for example "<tt>{ opaque, i32 }</tt>").
2053 Third, a concrete type is a type that is not an abstract type (e.g. "<tt>{ i32,
2059 <!-- ______________________________________________________________________ -->
2060 <div class="doc_subsubsection">
2061 <a name="BuildRecType">Basic Recursive Type Construction</a>
2064 <div class="doc_text">
2067 Because the most common question is "how do I build a recursive type with LLVM",
2068 we answer it now and explain it as we go. Here we include enough to cause this
2069 to be emitted to an output .ll file:
2072 <div class="doc_code">
2074 %mylist = type { %mylist*, i32 }
2079 To build this, use the following LLVM APIs:
2082 <div class="doc_code">
2084 // <i>Create the initial outer struct</i>
2085 <a href="#PATypeHolder">PATypeHolder</a> StructTy = OpaqueType::get();
2086 std::vector<const Type*> Elts;
2087 Elts.push_back(PointerType::get(StructTy));
2088 Elts.push_back(Type::Int32Ty);
2089 StructType *NewSTy = StructType::get(Elts);
2091 // <i>At this point, NewSTy = "{ opaque*, i32 }". Tell VMCore that</i>
2092 // <i>the struct and the opaque type are actually the same.</i>
2093 cast<OpaqueType>(StructTy.get())-><a href="#refineAbstractTypeTo">refineAbstractTypeTo</a>(NewSTy);
2095 // <i>NewSTy is potentially invalidated, but StructTy (a <a href="#PATypeHolder">PATypeHolder</a>) is</i>
2096 // <i>kept up-to-date</i>
2097 NewSTy = cast<StructType>(StructTy.get());
2099 // <i>Add a name for the type to the module symbol table (optional)</i>
2100 MyModule->addTypeName("mylist", NewSTy);
2105 This code shows the basic approach used to build recursive types: build a
2106 non-recursive type using 'opaque', then use type unification to close the cycle.
2107 The type unification step is performed by the <tt><a
2108 href="#refineAbstractTypeTo">refineAbstractTypeTo</a></tt> method, which is
2109 described next. After that, we describe the <a
2110 href="#PATypeHolder">PATypeHolder class</a>.
2115 <!-- ______________________________________________________________________ -->
2116 <div class="doc_subsubsection">
2117 <a name="refineAbstractTypeTo">The <tt>refineAbstractTypeTo</tt> method</a>
2120 <div class="doc_text">
2122 The <tt>refineAbstractTypeTo</tt> method starts the type unification process.
2123 While this method is actually a member of the DerivedType class, it is most
2124 often used on OpaqueType instances. Type unification is actually a recursive
2125 process. After unification, types can become structurally isomorphic to
2126 existing types, and all duplicates are deleted (to preserve pointer equality).
2130 In the example above, the OpaqueType object is definitely deleted.
2131 Additionally, if there is an "{ \2*, i32}" type already created in the system,
2132 the pointer and struct type created are <b>also</b> deleted. Obviously whenever
2133 a type is deleted, any "Type*" pointers in the program are invalidated. As
2134 such, it is safest to avoid having <i>any</i> "Type*" pointers to abstract types
2135 live across a call to <tt>refineAbstractTypeTo</tt> (note that non-abstract
2136 types can never move or be deleted). To deal with this, the <a
2137 href="#PATypeHolder">PATypeHolder</a> class is used to maintain a stable
2138 reference to a possibly refined type, and the <a
2139 href="#AbstractTypeUser">AbstractTypeUser</a> class is used to update more
2140 complex datastructures.
2145 <!-- ______________________________________________________________________ -->
2146 <div class="doc_subsubsection">
2147 <a name="PATypeHolder">The PATypeHolder Class</a>
2150 <div class="doc_text">
2152 PATypeHolder is a form of a "smart pointer" for Type objects. When VMCore
2153 happily goes about nuking types that become isomorphic to existing types, it
2154 automatically updates all PATypeHolder objects to point to the new type. In the
2155 example above, this allows the code to maintain a pointer to the resultant
2156 resolved recursive type, even though the Type*'s are potentially invalidated.
2160 PATypeHolder is an extremely light-weight object that uses a lazy union-find
2161 implementation to update pointers. For example the pointer from a Value to its
2162 Type is maintained by PATypeHolder objects.
2167 <!-- ______________________________________________________________________ -->
2168 <div class="doc_subsubsection">
2169 <a name="AbstractTypeUser">The AbstractTypeUser Class</a>
2172 <div class="doc_text">
2175 Some data structures need more to perform more complex updates when types get
2176 resolved. To support this, a class can derive from the AbstractTypeUser class.
2178 allows it to get callbacks when certain types are resolved. To register to get
2179 callbacks for a particular type, the DerivedType::{add/remove}AbstractTypeUser
2180 methods can be called on a type. Note that these methods only work for <i>
2181 abstract</i> types. Concrete types (those that do not include any opaque
2182 objects) can never be refined.
2187 <!-- ======================================================================= -->
2188 <div class="doc_subsection">
2189 <a name="SymbolTable">The <tt>ValueSymbolTable</tt> and
2190 <tt>TypeSymbolTable</tt> classes</a>
2193 <div class="doc_text">
2194 <p>The <tt><a href="http://llvm.org/doxygen/classllvm_1_1ValueSymbolTable.html">
2195 ValueSymbolTable</a></tt> class provides a symbol table that the <a
2196 href="#Function"><tt>Function</tt></a> and <a href="#Module">
2197 <tt>Module</tt></a> classes use for naming value definitions. The symbol table
2198 can provide a name for any <a href="#Value"><tt>Value</tt></a>.
2199 The <tt><a href="http://llvm.org/doxygen/classllvm_1_1TypeSymbolTable.html">
2200 TypeSymbolTable</a></tt> class is used by the <tt>Module</tt> class to store
2201 names for types.</p>
2203 <p>Note that the <tt>SymbolTable</tt> class should not be directly accessed
2204 by most clients. It should only be used when iteration over the symbol table
2205 names themselves are required, which is very special purpose. Note that not
2207 <a href="#Value">Value</a>s have names, and those without names (i.e. they have
2208 an empty name) do not exist in the symbol table.
2211 <p>These symbol tables support iteration over the values/types in the symbol
2212 table with <tt>begin/end/iterator</tt> and supports querying to see if a
2213 specific name is in the symbol table (with <tt>lookup</tt>). The
2214 <tt>ValueSymbolTable</tt> class exposes no public mutator methods, instead,
2215 simply call <tt>setName</tt> on a value, which will autoinsert it into the
2216 appropriate symbol table. For types, use the Module::addTypeName method to
2217 insert entries into the symbol table.</p>
2223 <!-- *********************************************************************** -->
2224 <div class="doc_section">
2225 <a name="coreclasses">The Core LLVM Class Hierarchy Reference </a>
2227 <!-- *********************************************************************** -->
2229 <div class="doc_text">
2230 <p><tt>#include "<a href="/doxygen/Type_8h-source.html">llvm/Type.h</a>"</tt>
2231 <br>doxygen info: <a href="/doxygen/classllvm_1_1Type.html">Type Class</a></p>
2233 <p>The Core LLVM classes are the primary means of representing the program
2234 being inspected or transformed. The core LLVM classes are defined in
2235 header files in the <tt>include/llvm/</tt> directory, and implemented in
2236 the <tt>lib/VMCore</tt> directory.</p>
2240 <!-- ======================================================================= -->
2241 <div class="doc_subsection">
2242 <a name="Type">The <tt>Type</tt> class and Derived Types</a>
2245 <div class="doc_text">
2247 <p><tt>Type</tt> is a superclass of all type classes. Every <tt>Value</tt> has
2248 a <tt>Type</tt>. <tt>Type</tt> cannot be instantiated directly but only
2249 through its subclasses. Certain primitive types (<tt>VoidType</tt>,
2250 <tt>LabelType</tt>, <tt>FloatType</tt> and <tt>DoubleType</tt>) have hidden
2251 subclasses. They are hidden because they offer no useful functionality beyond
2252 what the <tt>Type</tt> class offers except to distinguish themselves from
2253 other subclasses of <tt>Type</tt>.</p>
2254 <p>All other types are subclasses of <tt>DerivedType</tt>. Types can be
2255 named, but this is not a requirement. There exists exactly
2256 one instance of a given shape at any one time. This allows type equality to
2257 be performed with address equality of the Type Instance. That is, given two
2258 <tt>Type*</tt> values, the types are identical if the pointers are identical.
2262 <!-- _______________________________________________________________________ -->
2263 <div class="doc_subsubsection">
2264 <a name="m_Value">Important Public Methods</a>
2267 <div class="doc_text">
2270 <li><tt>bool isInteger() const</tt>: Returns true for any integer type.</li>
2272 <li><tt>bool isFloatingPoint()</tt>: Return true if this is one of the two
2273 floating point types.</li>
2275 <li><tt>bool isAbstract()</tt>: Return true if the type is abstract (contains
2276 an OpaqueType anywhere in its definition).</li>
2278 <li><tt>bool isSized()</tt>: Return true if the type has known size. Things
2279 that don't have a size are abstract types, labels and void.</li>
2284 <!-- _______________________________________________________________________ -->
2285 <div class="doc_subsubsection">
2286 <a name="m_Value">Important Derived Types</a>
2288 <div class="doc_text">
2290 <dt><tt>IntegerType</tt></dt>
2291 <dd>Subclass of DerivedType that represents integer types of any bit width.
2292 Any bit width between <tt>IntegerType::MIN_INT_BITS</tt> (1) and
2293 <tt>IntegerType::MAX_INT_BITS</tt> (~8 million) can be represented.
2295 <li><tt>static const IntegerType* get(unsigned NumBits)</tt>: get an integer
2296 type of a specific bit width.</li>
2297 <li><tt>unsigned getBitWidth() const</tt>: Get the bit width of an integer
2301 <dt><tt>SequentialType</tt></dt>
2302 <dd>This is subclassed by ArrayType and PointerType
2304 <li><tt>const Type * getElementType() const</tt>: Returns the type of each
2305 of the elements in the sequential type. </li>
2308 <dt><tt>ArrayType</tt></dt>
2309 <dd>This is a subclass of SequentialType and defines the interface for array
2312 <li><tt>unsigned getNumElements() const</tt>: Returns the number of
2313 elements in the array. </li>
2316 <dt><tt>PointerType</tt></dt>
2317 <dd>Subclass of SequentialType for pointer types.</dd>
2318 <dt><tt>VectorType</tt></dt>
2319 <dd>Subclass of SequentialType for vector types. A
2320 vector type is similar to an ArrayType but is distinguished because it is
2321 a first class type wherease ArrayType is not. Vector types are used for
2322 vector operations and are usually small vectors of of an integer or floating
2324 <dt><tt>StructType</tt></dt>
2325 <dd>Subclass of DerivedTypes for struct types.</dd>
2326 <dt><tt><a name="FunctionType">FunctionType</a></tt></dt>
2327 <dd>Subclass of DerivedTypes for function types.
2329 <li><tt>bool isVarArg() const</tt>: Returns true if its a vararg
2331 <li><tt> const Type * getReturnType() const</tt>: Returns the
2332 return type of the function.</li>
2333 <li><tt>const Type * getParamType (unsigned i)</tt>: Returns
2334 the type of the ith parameter.</li>
2335 <li><tt> const unsigned getNumParams() const</tt>: Returns the
2336 number of formal parameters.</li>
2339 <dt><tt>OpaqueType</tt></dt>
2340 <dd>Sublcass of DerivedType for abstract types. This class
2341 defines no content and is used as a placeholder for some other type. Note
2342 that OpaqueType is used (temporarily) during type resolution for forward
2343 references of types. Once the referenced type is resolved, the OpaqueType
2344 is replaced with the actual type. OpaqueType can also be used for data
2345 abstraction. At link time opaque types can be resolved to actual types
2346 of the same name.</dd>
2352 <!-- ======================================================================= -->
2353 <div class="doc_subsection">
2354 <a name="Module">The <tt>Module</tt> class</a>
2357 <div class="doc_text">
2360 href="/doxygen/Module_8h-source.html">llvm/Module.h</a>"</tt><br> doxygen info:
2361 <a href="/doxygen/classllvm_1_1Module.html">Module Class</a></p>
2363 <p>The <tt>Module</tt> class represents the top level structure present in LLVM
2364 programs. An LLVM module is effectively either a translation unit of the
2365 original program or a combination of several translation units merged by the
2366 linker. The <tt>Module</tt> class keeps track of a list of <a
2367 href="#Function"><tt>Function</tt></a>s, a list of <a
2368 href="#GlobalVariable"><tt>GlobalVariable</tt></a>s, and a <a
2369 href="#SymbolTable"><tt>SymbolTable</tt></a>. Additionally, it contains a few
2370 helpful member functions that try to make common operations easy.</p>
2374 <!-- _______________________________________________________________________ -->
2375 <div class="doc_subsubsection">
2376 <a name="m_Module">Important Public Members of the <tt>Module</tt> class</a>
2379 <div class="doc_text">
2382 <li><tt>Module::Module(std::string name = "")</tt></li>
2385 <p>Constructing a <a href="#Module">Module</a> is easy. You can optionally
2386 provide a name for it (probably based on the name of the translation unit).</p>
2389 <li><tt>Module::iterator</tt> - Typedef for function list iterator<br>
2390 <tt>Module::const_iterator</tt> - Typedef for const_iterator.<br>
2392 <tt>begin()</tt>, <tt>end()</tt>
2393 <tt>size()</tt>, <tt>empty()</tt>
2395 <p>These are forwarding methods that make it easy to access the contents of
2396 a <tt>Module</tt> object's <a href="#Function"><tt>Function</tt></a>
2399 <li><tt>Module::FunctionListType &getFunctionList()</tt>
2401 <p> Returns the list of <a href="#Function"><tt>Function</tt></a>s. This is
2402 necessary to use when you need to update the list or perform a complex
2403 action that doesn't have a forwarding method.</p>
2405 <p><!-- Global Variable --></p></li>
2411 <li><tt>Module::global_iterator</tt> - Typedef for global variable list iterator<br>
2413 <tt>Module::const_global_iterator</tt> - Typedef for const_iterator.<br>
2415 <tt>global_begin()</tt>, <tt>global_end()</tt>
2416 <tt>global_size()</tt>, <tt>global_empty()</tt>
2418 <p> These are forwarding methods that make it easy to access the contents of
2419 a <tt>Module</tt> object's <a
2420 href="#GlobalVariable"><tt>GlobalVariable</tt></a> list.</p></li>
2422 <li><tt>Module::GlobalListType &getGlobalList()</tt>
2424 <p>Returns the list of <a
2425 href="#GlobalVariable"><tt>GlobalVariable</tt></a>s. This is necessary to
2426 use when you need to update the list or perform a complex action that
2427 doesn't have a forwarding method.</p>
2429 <p><!-- Symbol table stuff --> </p></li>
2435 <li><tt><a href="#SymbolTable">SymbolTable</a> *getSymbolTable()</tt>
2437 <p>Return a reference to the <a href="#SymbolTable"><tt>SymbolTable</tt></a>
2438 for this <tt>Module</tt>.</p>
2440 <p><!-- Convenience methods --></p></li>
2446 <li><tt><a href="#Function">Function</a> *getFunction(const std::string
2447 &Name, const <a href="#FunctionType">FunctionType</a> *Ty)</tt>
2449 <p>Look up the specified function in the <tt>Module</tt> <a
2450 href="#SymbolTable"><tt>SymbolTable</tt></a>. If it does not exist, return
2451 <tt>null</tt>.</p></li>
2453 <li><tt><a href="#Function">Function</a> *getOrInsertFunction(const
2454 std::string &Name, const <a href="#FunctionType">FunctionType</a> *T)</tt>
2456 <p>Look up the specified function in the <tt>Module</tt> <a
2457 href="#SymbolTable"><tt>SymbolTable</tt></a>. If it does not exist, add an
2458 external declaration for the function and return it.</p></li>
2460 <li><tt>std::string getTypeName(const <a href="#Type">Type</a> *Ty)</tt>
2462 <p>If there is at least one entry in the <a
2463 href="#SymbolTable"><tt>SymbolTable</tt></a> for the specified <a
2464 href="#Type"><tt>Type</tt></a>, return it. Otherwise return the empty
2467 <li><tt>bool addTypeName(const std::string &Name, const <a
2468 href="#Type">Type</a> *Ty)</tt>
2470 <p>Insert an entry in the <a href="#SymbolTable"><tt>SymbolTable</tt></a>
2471 mapping <tt>Name</tt> to <tt>Ty</tt>. If there is already an entry for this
2472 name, true is returned and the <a
2473 href="#SymbolTable"><tt>SymbolTable</tt></a> is not modified.</p></li>
2479 <!-- ======================================================================= -->
2480 <div class="doc_subsection">
2481 <a name="Value">The <tt>Value</tt> class</a>
2484 <div class="doc_text">
2486 <p><tt>#include "<a href="/doxygen/Value_8h-source.html">llvm/Value.h</a>"</tt>
2488 doxygen info: <a href="/doxygen/classllvm_1_1Value.html">Value Class</a></p>
2490 <p>The <tt>Value</tt> class is the most important class in the LLVM Source
2491 base. It represents a typed value that may be used (among other things) as an
2492 operand to an instruction. There are many different types of <tt>Value</tt>s,
2493 such as <a href="#Constant"><tt>Constant</tt></a>s,<a
2494 href="#Argument"><tt>Argument</tt></a>s. Even <a
2495 href="#Instruction"><tt>Instruction</tt></a>s and <a
2496 href="#Function"><tt>Function</tt></a>s are <tt>Value</tt>s.</p>
2498 <p>A particular <tt>Value</tt> may be used many times in the LLVM representation
2499 for a program. For example, an incoming argument to a function (represented
2500 with an instance of the <a href="#Argument">Argument</a> class) is "used" by
2501 every instruction in the function that references the argument. To keep track
2502 of this relationship, the <tt>Value</tt> class keeps a list of all of the <a
2503 href="#User"><tt>User</tt></a>s that is using it (the <a
2504 href="#User"><tt>User</tt></a> class is a base class for all nodes in the LLVM
2505 graph that can refer to <tt>Value</tt>s). This use list is how LLVM represents
2506 def-use information in the program, and is accessible through the <tt>use_</tt>*
2507 methods, shown below.</p>
2509 <p>Because LLVM is a typed representation, every LLVM <tt>Value</tt> is typed,
2510 and this <a href="#Type">Type</a> is available through the <tt>getType()</tt>
2511 method. In addition, all LLVM values can be named. The "name" of the
2512 <tt>Value</tt> is a symbolic string printed in the LLVM code:</p>
2514 <div class="doc_code">
2516 %<b>foo</b> = add i32 1, 2
2520 <p><a name="nameWarning">The name of this instruction is "foo".</a> <b>NOTE</b>
2521 that the name of any value may be missing (an empty string), so names should
2522 <b>ONLY</b> be used for debugging (making the source code easier to read,
2523 debugging printouts), they should not be used to keep track of values or map
2524 between them. For this purpose, use a <tt>std::map</tt> of pointers to the
2525 <tt>Value</tt> itself instead.</p>
2527 <p>One important aspect of LLVM is that there is no distinction between an SSA
2528 variable and the operation that produces it. Because of this, any reference to
2529 the value produced by an instruction (or the value available as an incoming
2530 argument, for example) is represented as a direct pointer to the instance of
2532 represents this value. Although this may take some getting used to, it
2533 simplifies the representation and makes it easier to manipulate.</p>
2537 <!-- _______________________________________________________________________ -->
2538 <div class="doc_subsubsection">
2539 <a name="m_Value">Important Public Members of the <tt>Value</tt> class</a>
2542 <div class="doc_text">
2545 <li><tt>Value::use_iterator</tt> - Typedef for iterator over the
2547 <tt>Value::use_const_iterator</tt> - Typedef for const_iterator over
2549 <tt>unsigned use_size()</tt> - Returns the number of users of the
2551 <tt>bool use_empty()</tt> - Returns true if there are no users.<br>
2552 <tt>use_iterator use_begin()</tt> - Get an iterator to the start of
2554 <tt>use_iterator use_end()</tt> - Get an iterator to the end of the
2556 <tt><a href="#User">User</a> *use_back()</tt> - Returns the last
2557 element in the list.
2558 <p> These methods are the interface to access the def-use
2559 information in LLVM. As with all other iterators in LLVM, the naming
2560 conventions follow the conventions defined by the <a href="#stl">STL</a>.</p>
2562 <li><tt><a href="#Type">Type</a> *getType() const</tt>
2563 <p>This method returns the Type of the Value.</p>
2565 <li><tt>bool hasName() const</tt><br>
2566 <tt>std::string getName() const</tt><br>
2567 <tt>void setName(const std::string &Name)</tt>
2568 <p> This family of methods is used to access and assign a name to a <tt>Value</tt>,
2569 be aware of the <a href="#nameWarning">precaution above</a>.</p>
2571 <li><tt>void replaceAllUsesWith(Value *V)</tt>
2573 <p>This method traverses the use list of a <tt>Value</tt> changing all <a
2574 href="#User"><tt>User</tt>s</a> of the current value to refer to
2575 "<tt>V</tt>" instead. For example, if you detect that an instruction always
2576 produces a constant value (for example through constant folding), you can
2577 replace all uses of the instruction with the constant like this:</p>
2579 <div class="doc_code">
2581 Inst->replaceAllUsesWith(ConstVal);
2589 <!-- ======================================================================= -->
2590 <div class="doc_subsection">
2591 <a name="User">The <tt>User</tt> class</a>
2594 <div class="doc_text">
2597 <tt>#include "<a href="/doxygen/User_8h-source.html">llvm/User.h</a>"</tt><br>
2598 doxygen info: <a href="/doxygen/classllvm_1_1User.html">User Class</a><br>
2599 Superclass: <a href="#Value"><tt>Value</tt></a></p>
2601 <p>The <tt>User</tt> class is the common base class of all LLVM nodes that may
2602 refer to <a href="#Value"><tt>Value</tt></a>s. It exposes a list of "Operands"
2603 that are all of the <a href="#Value"><tt>Value</tt></a>s that the User is
2604 referring to. The <tt>User</tt> class itself is a subclass of
2607 <p>The operands of a <tt>User</tt> point directly to the LLVM <a
2608 href="#Value"><tt>Value</tt></a> that it refers to. Because LLVM uses Static
2609 Single Assignment (SSA) form, there can only be one definition referred to,
2610 allowing this direct connection. This connection provides the use-def
2611 information in LLVM.</p>
2615 <!-- _______________________________________________________________________ -->
2616 <div class="doc_subsubsection">
2617 <a name="m_User">Important Public Members of the <tt>User</tt> class</a>
2620 <div class="doc_text">
2622 <p>The <tt>User</tt> class exposes the operand list in two ways: through
2623 an index access interface and through an iterator based interface.</p>
2626 <li><tt>Value *getOperand(unsigned i)</tt><br>
2627 <tt>unsigned getNumOperands()</tt>
2628 <p> These two methods expose the operands of the <tt>User</tt> in a
2629 convenient form for direct access.</p></li>
2631 <li><tt>User::op_iterator</tt> - Typedef for iterator over the operand
2633 <tt>op_iterator op_begin()</tt> - Get an iterator to the start of
2634 the operand list.<br>
2635 <tt>op_iterator op_end()</tt> - Get an iterator to the end of the
2637 <p> Together, these methods make up the iterator based interface to
2638 the operands of a <tt>User</tt>.</p></li>
2643 <!-- ======================================================================= -->
2644 <div class="doc_subsection">
2645 <a name="Instruction">The <tt>Instruction</tt> class</a>
2648 <div class="doc_text">
2650 <p><tt>#include "</tt><tt><a
2651 href="/doxygen/Instruction_8h-source.html">llvm/Instruction.h</a>"</tt><br>
2652 doxygen info: <a href="/doxygen/classllvm_1_1Instruction.html">Instruction Class</a><br>
2653 Superclasses: <a href="#User"><tt>User</tt></a>, <a
2654 href="#Value"><tt>Value</tt></a></p>
2656 <p>The <tt>Instruction</tt> class is the common base class for all LLVM
2657 instructions. It provides only a few methods, but is a very commonly used
2658 class. The primary data tracked by the <tt>Instruction</tt> class itself is the
2659 opcode (instruction type) and the parent <a
2660 href="#BasicBlock"><tt>BasicBlock</tt></a> the <tt>Instruction</tt> is embedded
2661 into. To represent a specific type of instruction, one of many subclasses of
2662 <tt>Instruction</tt> are used.</p>
2664 <p> Because the <tt>Instruction</tt> class subclasses the <a
2665 href="#User"><tt>User</tt></a> class, its operands can be accessed in the same
2666 way as for other <a href="#User"><tt>User</tt></a>s (with the
2667 <tt>getOperand()</tt>/<tt>getNumOperands()</tt> and
2668 <tt>op_begin()</tt>/<tt>op_end()</tt> methods).</p> <p> An important file for
2669 the <tt>Instruction</tt> class is the <tt>llvm/Instruction.def</tt> file. This
2670 file contains some meta-data about the various different types of instructions
2671 in LLVM. It describes the enum values that are used as opcodes (for example
2672 <tt>Instruction::Add</tt> and <tt>Instruction::ICmp</tt>), as well as the
2673 concrete sub-classes of <tt>Instruction</tt> that implement the instruction (for
2674 example <tt><a href="#BinaryOperator">BinaryOperator</a></tt> and <tt><a
2675 href="#CmpInst">CmpInst</a></tt>). Unfortunately, the use of macros in
2676 this file confuses doxygen, so these enum values don't show up correctly in the
2677 <a href="/doxygen/classllvm_1_1Instruction.html">doxygen output</a>.</p>
2681 <!-- _______________________________________________________________________ -->
2682 <div class="doc_subsubsection">
2683 <a name="s_Instruction">Important Subclasses of the <tt>Instruction</tt>
2686 <div class="doc_text">
2688 <li><tt><a name="BinaryOperator">BinaryOperator</a></tt>
2689 <p>This subclasses represents all two operand instructions whose operands
2690 must be the same type, except for the comparison instructions.</p></li>
2691 <li><tt><a name="CastInst">CastInst</a></tt>
2692 <p>This subclass is the parent of the 12 casting instructions. It provides
2693 common operations on cast instructions.</p>
2694 <li><tt><a name="CmpInst">CmpInst</a></tt>
2695 <p>This subclass respresents the two comparison instructions,
2696 <a href="LangRef.html#i_icmp">ICmpInst</a> (integer opreands), and
2697 <a href="LangRef.html#i_fcmp">FCmpInst</a> (floating point operands).</p>
2698 <li><tt><a name="TerminatorInst">TerminatorInst</a></tt>
2699 <p>This subclass is the parent of all terminator instructions (those which
2700 can terminate a block).</p>
2704 <!-- _______________________________________________________________________ -->
2705 <div class="doc_subsubsection">
2706 <a name="m_Instruction">Important Public Members of the <tt>Instruction</tt>
2710 <div class="doc_text">
2713 <li><tt><a href="#BasicBlock">BasicBlock</a> *getParent()</tt>
2714 <p>Returns the <a href="#BasicBlock"><tt>BasicBlock</tt></a> that
2715 this <tt>Instruction</tt> is embedded into.</p></li>
2716 <li><tt>bool mayWriteToMemory()</tt>
2717 <p>Returns true if the instruction writes to memory, i.e. it is a
2718 <tt>call</tt>,<tt>free</tt>,<tt>invoke</tt>, or <tt>store</tt>.</p></li>
2719 <li><tt>unsigned getOpcode()</tt>
2720 <p>Returns the opcode for the <tt>Instruction</tt>.</p></li>
2721 <li><tt><a href="#Instruction">Instruction</a> *clone() const</tt>
2722 <p>Returns another instance of the specified instruction, identical
2723 in all ways to the original except that the instruction has no parent
2724 (ie it's not embedded into a <a href="#BasicBlock"><tt>BasicBlock</tt></a>),
2725 and it has no name</p></li>
2730 <!-- ======================================================================= -->
2731 <div class="doc_subsection">
2732 <a name="Constant">The <tt>Constant</tt> class and subclasses</a>
2735 <div class="doc_text">
2737 <p>Constant represents a base class for different types of constants. It
2738 is subclassed by ConstantInt, ConstantArray, etc. for representing
2739 the various types of Constants. <a href="#GlobalValue">GlobalValue</a> is also
2740 a subclass, which represents the address of a global variable or function.
2745 <!-- _______________________________________________________________________ -->
2746 <div class="doc_subsubsection">Important Subclasses of Constant </div>
2747 <div class="doc_text">
2749 <li>ConstantInt : This subclass of Constant represents an integer constant of
2752 <li><tt>const APInt& getValue() const</tt>: Returns the underlying
2753 value of this constant, an APInt value.</li>
2754 <li><tt>int64_t getSExtValue() const</tt>: Converts the underlying APInt
2755 value to an int64_t via sign extension. If the value (not the bit width)
2756 of the APInt is too large to fit in an int64_t, an assertion will result.
2757 For this reason, use of this method is discouraged.</li>
2758 <li><tt>uint64_t getZExtValue() const</tt>: Converts the underlying APInt
2759 value to a uint64_t via zero extension. IF the value (not the bit width)
2760 of the APInt is too large to fit in a uint64_t, an assertion will result.
2761 For this reason, use of this method is discouraged.</li>
2762 <li><tt>static ConstantInt* get(const APInt& Val)</tt>: Returns the
2763 ConstantInt object that represents the value provided by <tt>Val</tt>.
2764 The type is implied as the IntegerType that corresponds to the bit width
2765 of <tt>Val</tt>.</li>
2766 <li><tt>static ConstantInt* get(const Type *Ty, uint64_t Val)</tt>:
2767 Returns the ConstantInt object that represents the value provided by
2768 <tt>Val</tt> for integer type <tt>Ty</tt>.</li>
2771 <li>ConstantFP : This class represents a floating point constant.
2773 <li><tt>double getValue() const</tt>: Returns the underlying value of
2774 this constant. </li>
2777 <li>ConstantArray : This represents a constant array.
2779 <li><tt>const std::vector<Use> &getValues() const</tt>: Returns
2780 a vector of component constants that makeup this array. </li>
2783 <li>ConstantStruct : This represents a constant struct.
2785 <li><tt>const std::vector<Use> &getValues() const</tt>: Returns
2786 a vector of component constants that makeup this array. </li>
2789 <li>GlobalValue : This represents either a global variable or a function. In
2790 either case, the value is a constant fixed address (after linking).
2796 <!-- ======================================================================= -->
2797 <div class="doc_subsection">
2798 <a name="GlobalValue">The <tt>GlobalValue</tt> class</a>
2801 <div class="doc_text">
2804 href="/doxygen/GlobalValue_8h-source.html">llvm/GlobalValue.h</a>"</tt><br>
2805 doxygen info: <a href="/doxygen/classllvm_1_1GlobalValue.html">GlobalValue
2807 Superclasses: <a href="#Constant"><tt>Constant</tt></a>,
2808 <a href="#User"><tt>User</tt></a>, <a href="#Value"><tt>Value</tt></a></p>
2810 <p>Global values (<a href="#GlobalVariable"><tt>GlobalVariable</tt></a>s or <a
2811 href="#Function"><tt>Function</tt></a>s) are the only LLVM values that are
2812 visible in the bodies of all <a href="#Function"><tt>Function</tt></a>s.
2813 Because they are visible at global scope, they are also subject to linking with
2814 other globals defined in different translation units. To control the linking
2815 process, <tt>GlobalValue</tt>s know their linkage rules. Specifically,
2816 <tt>GlobalValue</tt>s know whether they have internal or external linkage, as
2817 defined by the <tt>LinkageTypes</tt> enumeration.</p>
2819 <p>If a <tt>GlobalValue</tt> has internal linkage (equivalent to being
2820 <tt>static</tt> in C), it is not visible to code outside the current translation
2821 unit, and does not participate in linking. If it has external linkage, it is
2822 visible to external code, and does participate in linking. In addition to
2823 linkage information, <tt>GlobalValue</tt>s keep track of which <a
2824 href="#Module"><tt>Module</tt></a> they are currently part of.</p>
2826 <p>Because <tt>GlobalValue</tt>s are memory objects, they are always referred to
2827 by their <b>address</b>. As such, the <a href="#Type"><tt>Type</tt></a> of a
2828 global is always a pointer to its contents. It is important to remember this
2829 when using the <tt>GetElementPtrInst</tt> instruction because this pointer must
2830 be dereferenced first. For example, if you have a <tt>GlobalVariable</tt> (a
2831 subclass of <tt>GlobalValue)</tt> that is an array of 24 ints, type <tt>[24 x
2832 i32]</tt>, then the <tt>GlobalVariable</tt> is a pointer to that array. Although
2833 the address of the first element of this array and the value of the
2834 <tt>GlobalVariable</tt> are the same, they have different types. The
2835 <tt>GlobalVariable</tt>'s type is <tt>[24 x i32]</tt>. The first element's type
2836 is <tt>i32.</tt> Because of this, accessing a global value requires you to
2837 dereference the pointer with <tt>GetElementPtrInst</tt> first, then its elements
2838 can be accessed. This is explained in the <a href="LangRef.html#globalvars">LLVM
2839 Language Reference Manual</a>.</p>
2843 <!-- _______________________________________________________________________ -->
2844 <div class="doc_subsubsection">
2845 <a name="m_GlobalValue">Important Public Members of the <tt>GlobalValue</tt>
2849 <div class="doc_text">
2852 <li><tt>bool hasInternalLinkage() const</tt><br>
2853 <tt>bool hasExternalLinkage() const</tt><br>
2854 <tt>void setInternalLinkage(bool HasInternalLinkage)</tt>
2855 <p> These methods manipulate the linkage characteristics of the <tt>GlobalValue</tt>.</p>
2858 <li><tt><a href="#Module">Module</a> *getParent()</tt>
2859 <p> This returns the <a href="#Module"><tt>Module</tt></a> that the
2860 GlobalValue is currently embedded into.</p></li>
2865 <!-- ======================================================================= -->
2866 <div class="doc_subsection">
2867 <a name="Function">The <tt>Function</tt> class</a>
2870 <div class="doc_text">
2873 href="/doxygen/Function_8h-source.html">llvm/Function.h</a>"</tt><br> doxygen
2874 info: <a href="/doxygen/classllvm_1_1Function.html">Function Class</a><br>
2875 Superclasses: <a href="#GlobalValue"><tt>GlobalValue</tt></a>,
2876 <a href="#Constant"><tt>Constant</tt></a>,
2877 <a href="#User"><tt>User</tt></a>,
2878 <a href="#Value"><tt>Value</tt></a></p>
2880 <p>The <tt>Function</tt> class represents a single procedure in LLVM. It is
2881 actually one of the more complex classes in the LLVM heirarchy because it must
2882 keep track of a large amount of data. The <tt>Function</tt> class keeps track
2883 of a list of <a href="#BasicBlock"><tt>BasicBlock</tt></a>s, a list of formal
2884 <a href="#Argument"><tt>Argument</tt></a>s, and a
2885 <a href="#SymbolTable"><tt>SymbolTable</tt></a>.</p>
2887 <p>The list of <a href="#BasicBlock"><tt>BasicBlock</tt></a>s is the most
2888 commonly used part of <tt>Function</tt> objects. The list imposes an implicit
2889 ordering of the blocks in the function, which indicate how the code will be
2890 layed out by the backend. Additionally, the first <a
2891 href="#BasicBlock"><tt>BasicBlock</tt></a> is the implicit entry node for the
2892 <tt>Function</tt>. It is not legal in LLVM to explicitly branch to this initial
2893 block. There are no implicit exit nodes, and in fact there may be multiple exit
2894 nodes from a single <tt>Function</tt>. If the <a
2895 href="#BasicBlock"><tt>BasicBlock</tt></a> list is empty, this indicates that
2896 the <tt>Function</tt> is actually a function declaration: the actual body of the
2897 function hasn't been linked in yet.</p>
2899 <p>In addition to a list of <a href="#BasicBlock"><tt>BasicBlock</tt></a>s, the
2900 <tt>Function</tt> class also keeps track of the list of formal <a
2901 href="#Argument"><tt>Argument</tt></a>s that the function receives. This
2902 container manages the lifetime of the <a href="#Argument"><tt>Argument</tt></a>
2903 nodes, just like the <a href="#BasicBlock"><tt>BasicBlock</tt></a> list does for
2904 the <a href="#BasicBlock"><tt>BasicBlock</tt></a>s.</p>
2906 <p>The <a href="#SymbolTable"><tt>SymbolTable</tt></a> is a very rarely used
2907 LLVM feature that is only used when you have to look up a value by name. Aside
2908 from that, the <a href="#SymbolTable"><tt>SymbolTable</tt></a> is used
2909 internally to make sure that there are not conflicts between the names of <a
2910 href="#Instruction"><tt>Instruction</tt></a>s, <a
2911 href="#BasicBlock"><tt>BasicBlock</tt></a>s, or <a
2912 href="#Argument"><tt>Argument</tt></a>s in the function body.</p>
2914 <p>Note that <tt>Function</tt> is a <a href="#GlobalValue">GlobalValue</a>
2915 and therefore also a <a href="#Constant">Constant</a>. The value of the function
2916 is its address (after linking) which is guaranteed to be constant.</p>
2919 <!-- _______________________________________________________________________ -->
2920 <div class="doc_subsubsection">
2921 <a name="m_Function">Important Public Members of the <tt>Function</tt>
2925 <div class="doc_text">
2928 <li><tt>Function(const </tt><tt><a href="#FunctionType">FunctionType</a>
2929 *Ty, LinkageTypes Linkage, const std::string &N = "", Module* Parent = 0)</tt>
2931 <p>Constructor used when you need to create new <tt>Function</tt>s to add
2932 the the program. The constructor must specify the type of the function to
2933 create and what type of linkage the function should have. The <a
2934 href="#FunctionType"><tt>FunctionType</tt></a> argument
2935 specifies the formal arguments and return value for the function. The same
2936 <a href="#FunctionType"><tt>FunctionType</tt></a> value can be used to
2937 create multiple functions. The <tt>Parent</tt> argument specifies the Module
2938 in which the function is defined. If this argument is provided, the function
2939 will automatically be inserted into that module's list of
2942 <li><tt>bool isExternal()</tt>
2944 <p>Return whether or not the <tt>Function</tt> has a body defined. If the
2945 function is "external", it does not have a body, and thus must be resolved
2946 by linking with a function defined in a different translation unit.</p></li>
2948 <li><tt>Function::iterator</tt> - Typedef for basic block list iterator<br>
2949 <tt>Function::const_iterator</tt> - Typedef for const_iterator.<br>
2951 <tt>begin()</tt>, <tt>end()</tt>
2952 <tt>size()</tt>, <tt>empty()</tt>
2954 <p>These are forwarding methods that make it easy to access the contents of
2955 a <tt>Function</tt> object's <a href="#BasicBlock"><tt>BasicBlock</tt></a>
2958 <li><tt>Function::BasicBlockListType &getBasicBlockList()</tt>
2960 <p>Returns the list of <a href="#BasicBlock"><tt>BasicBlock</tt></a>s. This
2961 is necessary to use when you need to update the list or perform a complex
2962 action that doesn't have a forwarding method.</p></li>
2964 <li><tt>Function::arg_iterator</tt> - Typedef for the argument list
2966 <tt>Function::const_arg_iterator</tt> - Typedef for const_iterator.<br>
2968 <tt>arg_begin()</tt>, <tt>arg_end()</tt>
2969 <tt>arg_size()</tt>, <tt>arg_empty()</tt>
2971 <p>These are forwarding methods that make it easy to access the contents of
2972 a <tt>Function</tt> object's <a href="#Argument"><tt>Argument</tt></a>
2975 <li><tt>Function::ArgumentListType &getArgumentList()</tt>
2977 <p>Returns the list of <a href="#Argument"><tt>Argument</tt></a>s. This is
2978 necessary to use when you need to update the list or perform a complex
2979 action that doesn't have a forwarding method.</p></li>
2981 <li><tt><a href="#BasicBlock">BasicBlock</a> &getEntryBlock()</tt>
2983 <p>Returns the entry <a href="#BasicBlock"><tt>BasicBlock</tt></a> for the
2984 function. Because the entry block for the function is always the first
2985 block, this returns the first block of the <tt>Function</tt>.</p></li>
2987 <li><tt><a href="#Type">Type</a> *getReturnType()</tt><br>
2988 <tt><a href="#FunctionType">FunctionType</a> *getFunctionType()</tt>
2990 <p>This traverses the <a href="#Type"><tt>Type</tt></a> of the
2991 <tt>Function</tt> and returns the return type of the function, or the <a
2992 href="#FunctionType"><tt>FunctionType</tt></a> of the actual
2995 <li><tt><a href="#SymbolTable">SymbolTable</a> *getSymbolTable()</tt>
2997 <p> Return a pointer to the <a href="#SymbolTable"><tt>SymbolTable</tt></a>
2998 for this <tt>Function</tt>.</p></li>
3003 <!-- ======================================================================= -->
3004 <div class="doc_subsection">
3005 <a name="GlobalVariable">The <tt>GlobalVariable</tt> class</a>
3008 <div class="doc_text">
3011 href="/doxygen/GlobalVariable_8h-source.html">llvm/GlobalVariable.h</a>"</tt>
3013 doxygen info: <a href="/doxygen/classllvm_1_1GlobalVariable.html">GlobalVariable
3015 Superclasses: <a href="#GlobalValue"><tt>GlobalValue</tt></a>,
3016 <a href="#Constant"><tt>Constant</tt></a>,
3017 <a href="#User"><tt>User</tt></a>,
3018 <a href="#Value"><tt>Value</tt></a></p>
3020 <p>Global variables are represented with the (suprise suprise)
3021 <tt>GlobalVariable</tt> class. Like functions, <tt>GlobalVariable</tt>s are also
3022 subclasses of <a href="#GlobalValue"><tt>GlobalValue</tt></a>, and as such are
3023 always referenced by their address (global values must live in memory, so their
3024 "name" refers to their constant address). See
3025 <a href="#GlobalValue"><tt>GlobalValue</tt></a> for more on this. Global
3026 variables may have an initial value (which must be a
3027 <a href="#Constant"><tt>Constant</tt></a>), and if they have an initializer,
3028 they may be marked as "constant" themselves (indicating that their contents
3029 never change at runtime).</p>
3032 <!-- _______________________________________________________________________ -->
3033 <div class="doc_subsubsection">
3034 <a name="m_GlobalVariable">Important Public Members of the
3035 <tt>GlobalVariable</tt> class</a>
3038 <div class="doc_text">
3041 <li><tt>GlobalVariable(const </tt><tt><a href="#Type">Type</a> *Ty, bool
3042 isConstant, LinkageTypes& Linkage, <a href="#Constant">Constant</a>
3043 *Initializer = 0, const std::string &Name = "", Module* Parent = 0)</tt>
3045 <p>Create a new global variable of the specified type. If
3046 <tt>isConstant</tt> is true then the global variable will be marked as
3047 unchanging for the program. The Linkage parameter specifies the type of
3048 linkage (internal, external, weak, linkonce, appending) for the variable. If
3049 the linkage is InternalLinkage, WeakLinkage, or LinkOnceLinkage, then
3050 the resultant global variable will have internal linkage. AppendingLinkage
3051 concatenates together all instances (in different translation units) of the
3052 variable into a single variable but is only applicable to arrays. See
3053 the <a href="LangRef.html#modulestructure">LLVM Language Reference</a> for
3054 further details on linkage types. Optionally an initializer, a name, and the
3055 module to put the variable into may be specified for the global variable as
3058 <li><tt>bool isConstant() const</tt>
3060 <p>Returns true if this is a global variable that is known not to
3061 be modified at runtime.</p></li>
3063 <li><tt>bool hasInitializer()</tt>
3065 <p>Returns true if this <tt>GlobalVariable</tt> has an intializer.</p></li>
3067 <li><tt><a href="#Constant">Constant</a> *getInitializer()</tt>
3069 <p>Returns the intial value for a <tt>GlobalVariable</tt>. It is not legal
3070 to call this method if there is no initializer.</p></li>
3076 <!-- ======================================================================= -->
3077 <div class="doc_subsection">
3078 <a name="BasicBlock">The <tt>BasicBlock</tt> class</a>
3081 <div class="doc_text">
3084 href="/doxygen/BasicBlock_8h-source.html">llvm/BasicBlock.h</a>"</tt><br>
3085 doxygen info: <a href="/doxygen/structllvm_1_1BasicBlock.html">BasicBlock
3087 Superclass: <a href="#Value"><tt>Value</tt></a></p>
3089 <p>This class represents a single entry multiple exit section of the code,
3090 commonly known as a basic block by the compiler community. The
3091 <tt>BasicBlock</tt> class maintains a list of <a
3092 href="#Instruction"><tt>Instruction</tt></a>s, which form the body of the block.
3093 Matching the language definition, the last element of this list of instructions
3094 is always a terminator instruction (a subclass of the <a
3095 href="#TerminatorInst"><tt>TerminatorInst</tt></a> class).</p>
3097 <p>In addition to tracking the list of instructions that make up the block, the
3098 <tt>BasicBlock</tt> class also keeps track of the <a
3099 href="#Function"><tt>Function</tt></a> that it is embedded into.</p>
3101 <p>Note that <tt>BasicBlock</tt>s themselves are <a
3102 href="#Value"><tt>Value</tt></a>s, because they are referenced by instructions
3103 like branches and can go in the switch tables. <tt>BasicBlock</tt>s have type
3108 <!-- _______________________________________________________________________ -->
3109 <div class="doc_subsubsection">
3110 <a name="m_BasicBlock">Important Public Members of the <tt>BasicBlock</tt>
3114 <div class="doc_text">
3117 <li><tt>BasicBlock(const std::string &Name = "", </tt><tt><a
3118 href="#Function">Function</a> *Parent = 0)</tt>
3120 <p>The <tt>BasicBlock</tt> constructor is used to create new basic blocks for
3121 insertion into a function. The constructor optionally takes a name for the new
3122 block, and a <a href="#Function"><tt>Function</tt></a> to insert it into. If
3123 the <tt>Parent</tt> parameter is specified, the new <tt>BasicBlock</tt> is
3124 automatically inserted at the end of the specified <a
3125 href="#Function"><tt>Function</tt></a>, if not specified, the BasicBlock must be
3126 manually inserted into the <a href="#Function"><tt>Function</tt></a>.</p></li>
3128 <li><tt>BasicBlock::iterator</tt> - Typedef for instruction list iterator<br>
3129 <tt>BasicBlock::const_iterator</tt> - Typedef for const_iterator.<br>
3130 <tt>begin()</tt>, <tt>end()</tt>, <tt>front()</tt>, <tt>back()</tt>,
3131 <tt>size()</tt>, <tt>empty()</tt>
3132 STL-style functions for accessing the instruction list.
3134 <p>These methods and typedefs are forwarding functions that have the same
3135 semantics as the standard library methods of the same names. These methods
3136 expose the underlying instruction list of a basic block in a way that is easy to
3137 manipulate. To get the full complement of container operations (including
3138 operations to update the list), you must use the <tt>getInstList()</tt>
3141 <li><tt>BasicBlock::InstListType &getInstList()</tt>
3143 <p>This method is used to get access to the underlying container that actually
3144 holds the Instructions. This method must be used when there isn't a forwarding
3145 function in the <tt>BasicBlock</tt> class for the operation that you would like
3146 to perform. Because there are no forwarding functions for "updating"
3147 operations, you need to use this if you want to update the contents of a
3148 <tt>BasicBlock</tt>.</p></li>
3150 <li><tt><a href="#Function">Function</a> *getParent()</tt>
3152 <p> Returns a pointer to <a href="#Function"><tt>Function</tt></a> the block is
3153 embedded into, or a null pointer if it is homeless.</p></li>
3155 <li><tt><a href="#TerminatorInst">TerminatorInst</a> *getTerminator()</tt>
3157 <p> Returns a pointer to the terminator instruction that appears at the end of
3158 the <tt>BasicBlock</tt>. If there is no terminator instruction, or if the last
3159 instruction in the block is not a terminator, then a null pointer is
3167 <!-- ======================================================================= -->
3168 <div class="doc_subsection">
3169 <a name="Argument">The <tt>Argument</tt> class</a>
3172 <div class="doc_text">
3174 <p>This subclass of Value defines the interface for incoming formal
3175 arguments to a function. A Function maintains a list of its formal
3176 arguments. An argument has a pointer to the parent Function.</p>
3180 <!-- *********************************************************************** -->
3183 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
3184 src="http://jigsaw.w3.org/css-validator/images/vcss" alt="Valid CSS!"></a>
3185 <a href="http://validator.w3.org/check/referer"><img
3186 src="http://www.w3.org/Icons/valid-html401" alt="Valid HTML 4.01!" /></a>
3188 <a href="mailto:dhurjati@cs.uiuc.edu">Dinakar Dhurjati</a> and
3189 <a href="mailto:sabre@nondot.org">Chris Lattner</a><br>
3190 <a href="http://llvm.org">The LLVM Compiler Infrastructure</a><br>
3191 Last modified: $Date$