1 <!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN">
2 <html><head><title>LLVM Programmer's Manual</title></head>
6 <table width="100%" bgcolor="#330077" border=0 cellpadding=4 cellspacing=0>
7 <tr><td> <font size=+3 color="#EEEEFF" face="Georgia,Palatino,Times,Roman"><b>LLVM Programmer's Manual</b></font></td>
11 <li><a href="#introduction">Introduction</a>
12 <li><a href="#general">General Information</a>
14 <li><a href="#stl">The C++ Standard Template Library</a>
15 <li><a href="#isa">The <tt>isa<></tt>, <tt>cast<></tt> and
16 <tt>dyn_cast<></tt> templates</a>
18 <li><a href="#common">Helpful Hints for Common Operations</a>
20 <li><a href="#inspection">Basic Inspection and Traversal Routines</a>
22 <li><a href="#iterate_function">Iterating over the <tt>BasicBlock</tt>s
23 in a <tt>Function</tt></a>
24 <li><a href="#iterate_basicblock">Iterating over the <tt>Instruction</tt>s
25 in a <tt>BasicBlock</tt></a>
26 <li><a href="#iterate_institer">Iterating over the <tt>Instruction</tt>s
27 in a <tt>Function</tt></a>
28 <li><a href="#iterate_convert">Turning an iterator into a class
30 <li><a href="#iterate_complex">Finding call sites: a more complex
32 <li><a href="#iterate_chains">Iterating over def-use & use-def
35 <li><a href="#simplechanges">Making simple changes</a>
37 <li><a href="#schanges_creating">Creating and inserting new
38 <tt>Instruction</tt>s</a>
39 <li><a href="#schanges_deleting">Deleting
40 <tt>Instruction</tt>s</a>
41 <li><a href="#schanges_replacing">Replacing an
42 <tt>Instruction</tt> with another <tt>Value</tt></a>
45 <li>Working with the Control Flow Graph
47 <li>Accessing predecessors and successors of a <tt>BasicBlock</tt>
53 <li>The general graph API
54 <li>The <tt>InstVisitor</tt> template
56 <li>The <tt>Statistic</tt> template
60 <li>Useful related topics
62 <li>The <tt>-time-passes</tt> option
63 <li>How to use the LLVM Makefile system
64 <li>How to write a regression test
69 <li><a href="#coreclasses">The Core LLVM Class Hierarchy Reference</a>
71 <li><a href="#Value">The <tt>Value</tt> class</a>
73 <li><a href="#User">The <tt>User</tt> class</a>
75 <li><a href="#Instruction">The <tt>Instruction</tt> class</a>
79 <li><a href="#GlobalValue">The <tt>GlobalValue</tt> class</a>
81 <li><a href="#BasicBlock">The <tt>BasicBlock</tt> class</a>
82 <li><a href="#Function">The <tt>Function</tt> class</a>
83 <li><a href="#GlobalVariable">The <tt>GlobalVariable</tt> class</a>
85 <li><a href="#Module">The <tt>Module</tt> class</a>
86 <li><a href="#Constant">The <tt>Constant</tt> class</a>
92 <li><a href="#Type">The <tt>Type</tt> class</a>
93 <li><a href="#Argument">The <tt>Argument</tt> class</a>
95 <li>The <tt>SymbolTable</tt> class
96 <li>The <tt>ilist</tt> and <tt>iplist</tt> classes
98 <li>Creating, inserting, moving and deleting from LLVM lists
100 <li>Important iterator invalidation semantics to be aware of
103 <p><b>Written by <a href="mailto:sabre@nondot.org">Chris Lattner</a>,
104 <a href="mailto:dhurjati@cs.uiuc.edu">Dinakar Dhurjati</a>, and
105 <a href="mailto:jstanley@cs.uiuc.edu">Joel Stanley</a></b><p>
109 <!-- *********************************************************************** -->
110 <table width="100%" bgcolor="#330077" border=0 cellpadding=4 cellspacing=0>
111 <tr><td align=center><font color="#EEEEFF" size=+2 face="Georgia,Palatino"><b>
112 <a name="introduction">Introduction
113 </b></font></td></tr></table><ul>
114 <!-- *********************************************************************** -->
116 This document is meant to highlight some of the important classes and interfaces
117 available in the LLVM source-base. This manual is not intended to explain what
118 LLVM is, how it works, and what LLVM code looks like. It assumes that you know
119 the basics of LLVM and are interested in writing transformations or otherwise
120 analyzing or manipulating the code.<p>
122 This document should get you oriented so that you can find your way in the
123 continuously growing source code that makes up the LLVM infrastructure. Note
124 that this manual is not intended to serve as a replacement for reading the
125 source code, so if you think there should be a method in one of these classes to
126 do something, but it's not listed, check the source. Links to the <a
127 href="/doxygen/">doxygen</a> sources are provided to make this as easy as
130 The first section of this document describes general information that is useful
131 to know when working in the LLVM infrastructure, and the second describes the
132 Core LLVM classes. In the future this manual will be extended with information
133 describing how to use extension libraries, such as dominator information, CFG
134 traversal routines, and useful utilities like the <tt><a
135 href="/doxygen/InstVisitor_8h-source.html">InstVisitor</a></tt> template.<p>
138 <!-- *********************************************************************** -->
139 </ul><table width="100%" bgcolor="#330077" border=0 cellpadding=4 cellspacing=0>
140 <tr><td align=center><font color="#EEEEFF" size=+2 face="Georgia,Palatino"><b>
141 <a name="general">General Information
142 </b></font></td></tr></table><ul>
143 <!-- *********************************************************************** -->
145 This section contains general information that is useful if you are working in
146 the LLVM source-base, but that isn't specific to any particular API.<p>
149 <!-- ======================================================================= -->
150 </ul><table width="100%" bgcolor="#441188" border=0 cellpadding=4 cellspacing=0>
151 <tr><td> </td><td width="100%">
152 <font color="#EEEEFF" face="Georgia,Palatino"><b>
153 <a name="stl">The C++ Standard Template Library</a>
154 </b></font></td></tr></table><ul>
156 LLVM makes heavy use of the C++ Standard Template Library (STL), perhaps much
157 more than you are used to, or have seen before. Because of this, you might want
158 to do a little background reading in the techniques used and capabilities of the
159 library. There are many good pages that discuss the STL, and several books on
160 the subject that you can get, so it will not be discussed in this document.<p>
162 Here are some useful links:<p>
164 <li><a href="http://www.dinkumware.com/htm_cpl/index.html">Dinkumware C++
165 Library reference</a> - an excellent reference for the STL and other parts of
166 the standard C++ library.<br>
168 <li><a href="http://www.parashift.com/c++-faq-lite/">C++ Frequently Asked
171 <li><a href="http://www.sgi.com/tech/stl/">SGI's STL Programmer's Guide</a> -
173 href="http://www.sgi.com/tech/stl/stl_introduction.html">Introduction to the
176 <li><a href="http://www.research.att.com/~bs/C++.html">Bjarne Stroustrup's C++
181 You are also encouraged to take a look at the <a
182 href="CodingStandards.html">LLVM Coding Standards</a> guide which focuses on how
183 to write maintainable code more than where to put your curly braces.<p>
186 <!-- ======================================================================= -->
187 </ul><table width="100%" bgcolor="#441188" border=0 cellpadding=4 cellspacing=0>
188 <tr><td> </td><td width="100%">
189 <font color="#EEEEFF" face="Georgia,Palatino"><b>
190 <a name="isa">The isa<>, cast<> and dyn_cast<> templates</a>
191 </b></font></td></tr></table><ul>
193 The LLVM source-base makes extensive use of a custom form of RTTI. These
194 templates have many similarities to the C++ <tt>dynamic_cast<></tt>
195 operator, but they don't have some drawbacks (primarily stemming from the fact
196 that <tt>dynamic_cast<></tt> only works on classes that have a v-table).
197 Because they are used so often, you must know what they do and how they work.
198 All of these templates are defined in the <a
199 href="/doxygen/Casting_8h-source.html"><tt>Support/Casting.h</tt></a> file (note
200 that you very rarely have to include this file directly).<p>
204 <dt><tt>isa<></tt>:
206 <dd>The <tt>isa<></tt> operator works exactly like the Java
207 "<tt>instanceof</tt>" operator. It returns true or false depending on whether a
208 reference or pointer points to an instance of the specified class. This can be
209 very useful for constraint checking of various sorts (example below).<p>
212 <dt><tt>cast<></tt>:
214 <dd>The <tt>cast<></tt> operator is a "checked cast" operation. It
215 converts a pointer or reference from a base class to a derived cast, causing an
216 assertion failure if it is not really an instance of the right type. This
217 should be used in cases where you have some information that makes you believe
218 that something is of the right type. An example of the <tt>isa<></tt> and
219 <tt>cast<></tt> template is:<p>
222 static bool isLoopInvariant(const <a href="#Value">Value</a> *V, const Loop *L) {
223 if (isa<<a href="#Constant">Constant</a>>(V) || isa<<a href="#Argument">Argument</a>>(V) || isa<<a href="#GlobalValue">GlobalValue</a>>(V))
226 <i>// Otherwise, it must be an instruction...</i>
227 return !L->contains(cast<<a href="#Instruction">Instruction</a>>(V)->getParent());
230 Note that you should <b>not</b> use an <tt>isa<></tt> test followed by a
231 <tt>cast<></tt>, for that use the <tt>dyn_cast<></tt> operator.<p>
234 <dt><tt>dyn_cast<></tt>:
236 <dd>The <tt>dyn_cast<></tt> operator is a "checking cast" operation. It
237 checks to see if the operand is of the specified type, and if so, returns a
238 pointer to it (this operator does not work with references). If the operand is
239 not of the correct type, a null pointer is returned. Thus, this works very much
240 like the <tt>dynamic_cast</tt> operator in C++, and should be used in the same
241 circumstances. Typically, the <tt>dyn_cast<></tt> operator is used in an
242 <tt>if</tt> statement or some other flow control statement like this:<p>
245 if (<a href="#AllocationInst">AllocationInst</a> *AI = dyn_cast<<a href="#AllocationInst">AllocationInst</a>>(Val)) {
250 This form of the <tt>if</tt> statement effectively combines together a call to
251 <tt>isa<></tt> and a call to <tt>cast<></tt> into one statement,
252 which is very convenient.<p>
254 Another common example is:<p>
257 <i>// Loop over all of the phi nodes in a basic block</i>
258 BasicBlock::iterator BBI = BB->begin();
259 for (; <a href="#PhiNode">PHINode</a> *PN = dyn_cast<<a href="#PHINode">PHINode</a>>(&*BBI); ++BBI)
263 Note that the <tt>dyn_cast<></tt> operator, like C++'s
264 <tt>dynamic_cast</tt> or Java's <tt>instanceof</tt> operator, can be abused. In
265 particular you should not use big chained <tt>if/then/else</tt> blocks to check
266 for lots of different variants of classes. If you find yourself wanting to do
267 this, it is much cleaner and more efficient to use the InstVisitor class to
268 dispatch over the instruction type directly.<p>
271 <dt><tt>cast_or_null<></tt>:
273 <dd>The <tt>cast_or_null<></tt> operator works just like the
274 <tt>cast<></tt> operator, except that it allows for a null pointer as an
275 argument (which it then propagates). This can sometimes be useful, allowing you
276 to combine several null checks into one.<p>
279 <dt><tt>dyn_cast_or_null<></tt>:
281 <dd>The <tt>dyn_cast_or_null<></tt> operator works just like the
282 <tt>dyn_cast<></tt> operator, except that it allows for a null pointer as
283 an argument (which it then propagates). This can sometimes be useful, allowing
284 you to combine several null checks into one.<p>
288 These five templates can be used with any classes, whether they have a v-table
289 or not. To add support for these templates, you simply need to add
290 <tt>classof</tt> static methods to the class you are interested casting to.
291 Describing this is currently outside the scope of this document, but there are
292 lots of examples in the LLVM source base.<p>
296 <!-- *********************************************************************** -->
297 </ul><table width="100%" bgcolor="#330077" border=0 cellpadding=4 cellspacing=0>
298 <tr><td align=center><font color="#EEEEFF" size=+2 face="Georgia,Palatino"><b>
299 <a name="common">Helpful Hints for Common Operations
300 </b></font></td></tr></table><ul>
301 <!-- *********************************************************************** -->
303 This section describes how to perform some very simple transformations of LLVM
304 code. This is meant to give examples of common idioms used, showing the
305 practical side of LLVM transformations.<p>
307 Because this is a "how-to" section, you should also read about the main classes
308 that you will be working with. The <a href="#coreclasses">Core LLVM Class
309 Hierarchy Reference</a> contains details and descriptions of the main classes
310 that you should know about.<p>
312 <!-- NOTE: this section should be heavy on example code -->
315 <!-- ======================================================================= -->
316 </ul><table width="100%" bgcolor="#441188" border=0 cellpadding=4 cellspacing=0>
317 <tr><td> </td><td width="100%">
318 <font color="#EEEEFF" face="Georgia,Palatino"><b>
319 <a name="inspection">Basic Inspection and Traversal Routines</a>
320 </b></font></td></tr></table><ul>
322 The LLVM compiler infrastructure have many different data structures that may be
323 traversed. Following the example of the C++ standard template library, the
324 techniques used to traverse these various data structures are all basically the
325 same. For a enumerable sequence of values, the <tt>XXXbegin()</tt> function (or
326 method) returns an iterator to the start of the sequence, the <tt>XXXend()</tt>
327 function returns an iterator pointing to one past the last valid element of the
328 sequence, and there is some <tt>XXXiterator</tt> data type that is common
329 between the two operations.<p>
331 Because the pattern for iteration is common across many different aspects of the
332 program representation, the standard template library algorithms may be used on
333 them, and it is easier to remember how to iterate. First we show a few common
334 examples of the data structures that need to be traversed. Other data
335 structures are traversed in very similar ways.<p>
338 <!-- _______________________________________________________________________ -->
339 </ul><h4><a name="iterate_function"><hr size=0>Iterating over the <a
340 href="#BasicBlock"><tt>BasicBlock</tt></a>s in a <a
341 href="#Function"><tt>Function</tt></a> </h4><ul>
343 It's quite common to have a <tt>Function</tt> instance that you'd like
344 to transform in some way; in particular, you'd like to manipulate its
345 <tt>BasicBlock</tt>s. To facilitate this, you'll need to iterate over
346 all of the <tt>BasicBlock</tt>s that constitute the <tt>Function</tt>.
347 The following is an example that prints the name of a
348 <tt>BasicBlock</tt> and the number of <tt>Instruction</tt>s it
352 // func is a pointer to a Function instance
353 for(Function::iterator i = func->begin(), e = func->end(); i != e; ++i) {
355 // print out the name of the basic block if it has one, and then the
356 // number of instructions that it contains
358 cerr << "Basic block (name=" << i->getName() << ") has "
359 << i->size() << " instructions.\n";
363 Note that i can be used as if it were a pointer for the purposes of
364 invoking member functions of the <tt>Instruction</tt> class. This is
365 because the indirection operator is overloaded for the iterator
366 classes. In the above code, the expression <tt>i->size()</tt> is
367 exactly equivalent to <tt>(*i).size()</tt> just like you'd expect.
369 <!-- _______________________________________________________________________ -->
370 </ul><h4><a name="iterate_basicblock"><hr size=0>Iterating over the <a
371 href="#Instruction"><tt>Instruction</tt></a>s in a <a
372 href="#BasicBlock"><tt>BasicBlock</tt></a> </h4><ul>
374 Just like when dealing with <tt>BasicBlock</tt>s in
375 <tt>Function</tt>s, it's easy to iterate over the individual
376 instructions that make up <tt>BasicBlock</tt>s. Here's a code snippet
377 that prints out each instruction in a <tt>BasicBlock</tt>:
380 // blk is a pointer to a BasicBlock instance
381 for(BasicBlock::iterator i = blk->begin(), e = blk->end(); i != e; ++i)
382 // the next statement works since operator<<(ostream&,...)
383 // is overloaded for Instruction&
384 cerr << *i << "\n";
387 However, this isn't really the best way to print out the contents of a
388 <tt>BasicBlock</tt>! Since the ostream operators are overloaded for
389 virtually anything you'll care about, you could have just invoked the
390 print routine on the basic block itself: <tt>cerr << *blk <<
393 Note that currently operator<< is implemented for <tt>Value*</tt>, so it
394 will print out the contents of the pointer, instead of
395 the pointer value you might expect. This is a deprecated interface that will
396 be removed in the future, so it's best not to depend on it. To print out the
397 pointer value for now, you must cast to <tt>void*</tt>.<p>
400 <!-- _______________________________________________________________________ -->
401 </ul><h4><a name="iterate_institer"><hr size=0>Iterating over the <a
402 href="#Instruction"><tt>Instruction</tt></a>s in a <a
403 href="#Function"><tt>Function</tt></a></h4><ul>
405 If you're finding that you commonly iterate over a <tt>Function</tt>'s
406 <tt>BasicBlock</tt>s and then that <tt>BasicBlock</tt>'s
407 <tt>Instruction</tt>s, <tt>InstIterator</tt> should be used instead.
408 You'll need to include <a href="/doxygen/InstIterator_8h-source.html"><tt>llvm/Support/InstIterator.h</tt></a>, and then
409 instantiate <tt>InstIterator</tt>s explicitly in your code. Here's a
410 small example that shows how to dump all instructions in a function to
411 stderr (<b>Note:</b> Dereferencing an <tt>InstIterator</tt> yields an
412 <tt>Instruction*</tt>, <i>not</i> an <tt>Instruction&</tt>!):
415 #include "<a href="/doxygen/InstIterator_8h-source.html">llvm/Support/InstIterator.h</a>"
417 // Suppose F is a ptr to a function
418 for(inst_iterator i = inst_begin(F), e = inst_end(F); i != e; ++i)
419 cerr << **i << "\n";
422 Easy, isn't it? You can also use <tt>InstIterator</tt>s to fill a
423 worklist with its initial contents. For example, if you wanted to
424 initialize a worklist to contain all instructions in a
425 <tt>Function</tt> F, all you would need to do is something like:
428 std::set<Instruction*> worklist;
429 worklist.insert(inst_begin(F), inst_end(F));
432 The STL set <tt>worklist</tt> would now contain all instructions in
433 the <tt>Function</tt> pointed to by F.
435 <!-- _______________________________________________________________________ -->
436 </ul><h4><a name="iterate_convert"><hr size=0>Turning an iterator into a class
437 pointer (and vice-versa) </h4><ul>
439 Sometimes, it'll be useful to grab a reference (or pointer) to a class
440 instance when all you've got at hand is an iterator. Well, extracting
441 a reference or a pointer from an iterator is very straightforward.
442 Assuming that <tt>i</tt> is a <tt>BasicBlock::iterator</tt> and
443 <tt>j</tt> is a <tt>BasicBlock::const_iterator</tt>:
446 Instruction& inst = *i; // grab reference to instruction reference
447 Instruction* pinst = &*i; // grab pointer to instruction reference
448 const Instruction& inst = *j;
450 However, the iterators you'll be working with in the LLVM framework
451 are special: they will automatically convert to a ptr-to-instance type
452 whenever they need to. Instead of dereferencing the iterator and then
453 taking the address of the result, you can simply assign the iterator
454 to the proper pointer type and you get the dereference and address-of
455 operation as a result of the assignment (behind the scenes, this is a
456 result of overloading casting mechanisms). Thus the last line of the
459 <pre>Instruction* pinst = &*i;</pre>
461 is semantically equivalent to
463 <pre>Instruction* pinst = i;</pre>
465 <b>Caveat emptor</b>: The above syntax works <i>only</i> when you're <i>not</i>
466 working with <tt>dyn_cast</tt>. The template definition of <tt><a
467 href="#isa">dyn_cast</a></tt> isn't implemented to handle this yet, so you'll
468 still need the following in order for things to work properly:
471 BasicBlock::iterator bbi = ...;
472 <a href="#BranchInst">BranchInst</a>* b = <a href="#isa">dyn_cast</a><<a href="#BranchInst">BranchInst</a>>(&*bbi);
475 It's also possible to turn a class pointer into the corresponding
476 iterator. Usually, this conversion is quite inexpensive. The
477 following code snippet illustrates use of the conversion constructors
478 provided by LLVM iterators. By using these, you can explicitly grab
479 the iterator of something without actually obtaining it via iteration
483 void printNextInstruction(Instruction* inst) {
484 BasicBlock::iterator it(inst);
485 ++it; // after this line, it refers to the instruction after *inst.
486 if(it != inst->getParent()->end()) cerr << *it << "\n";
489 Of course, this example is strictly pedagogical, because it'd be much
490 better to explicitly grab the next instruction directly from inst.
493 <!--_______________________________________________________________________-->
494 </ul><h4><a name="iterate_complex"><hr size=0>Finding call sites: a slightly
495 more complex example </h4><ul>
497 Say that you're writing a FunctionPass and would like to count all the
498 locations in the entire module (that is, across every
499 <tt>Function</tt>) where a certain function (i.e. some
500 <tt>Function</tt>*) already in scope. As you'll learn later, you may
501 want to use an <tt>InstVisitor</tt> to accomplish this in a much more
502 straightforward manner, but this example will allow us to explore how
503 you'd do it if you didn't have <tt>InstVisitor</tt> around. In
504 pseudocode, this is what we want to do:
507 initialize callCounter to zero
508 for each Function f in the Module
509 for each BasicBlock b in f
510 for each Instruction i in b
511 if(i is a CallInst and calls the given function)
512 increment callCounter
515 And the actual code is (remember, since we're writing a
516 <tt>FunctionPass</tt>, our <tt>FunctionPass</tt>-derived class simply
517 has to override the <tt>runOnFunction</tt> method...):
520 Function* targetFunc = ...;
522 class OurFunctionPass : public FunctionPass {
524 OurFunctionPass(): callCounter(0) { }
526 virtual runOnFunction(Function& F) {
527 for(Function::iterator b = F.begin(), be = F.end(); b != be; ++b) {
528 for(BasicBlock::iterator i = b->begin(); ie = b->end(); i != ie; ++i) {
529 if (<a href="#CallInst">CallInst</a>* callInst = <a href="#isa">dyn_cast</a><<a href="#CallInst">CallInst</a>>(&*inst)) {
530 // we know we've encountered a call instruction, so we
531 // need to determine if it's a call to the
532 // function pointed to by m_func or not.
534 if(callInst->getCalledFunction() == targetFunc)
541 unsigned callCounter;
545 <!--_______________________________________________________________________-->
546 </ul><h4><a name="iterate_chains"><hr size=0>Iterating over def-use &
547 use-def chains</h4><ul>
549 Frequently, we might have an instance of the <a
550 href="/doxygen/classValue.html">Value Class</a> and we want to
551 determine which <tt>User</tt>s use the <tt>Value</tt>. The list of
552 all <tt>User</tt>s of a particular <tt>Value</tt> is called a
553 <i>def-use</i> chain. For example, let's say we have a
554 <tt>Function*</tt> named <tt>F</tt> to a particular function
555 <tt>foo</tt>. Finding all of the instructions that <i>use</i>
556 <tt>foo</tt> is as simple as iterating over the <i>def-use</i> chain of
562 for(Value::use_iterator i = F->use_begin(), e = F->use_end(); i != e; ++i) {
563 if(Instruction* i = dyn_cast<Instruction>(*i)) {
564 cerr << "F is used in instruction:\n\t";
565 cerr << *i << "\n";
570 Alternately, it's common to have an instance of the <a
571 href="/doxygen/classUser.html">User Class</a> and need to know what
572 <tt>Value</tt>s are used by it. The list of all <tt>Value</tt>s used
573 by a <tt>User</tt> is known as a <i>use-def</i> chain. Instances of
574 class <tt>Instruction</tt> are common <tt>User</tt>s, so we might want
575 to iterate over all of the values that a particular instruction uses
576 (that is, the operands of the particular <tt>Instruction</tt>):
579 Instruction* pi = ...;
581 for(User::op_iterator i = pi->op_begin(), e = pi->op_end(); i != e; ++i) {
589 def-use chains ("finding all users of"): Value::use_begin/use_end
590 use-def chains ("finding all values used"): User::op_begin/op_end [op=operand]
593 <!-- ======================================================================= -->
594 </ul><table width="100%" bgcolor="#441188" border=0 cellpadding=4 cellspacing=0>
595 <tr><td> </td><td width="100%">
596 <font color="#EEEEFF" face="Georgia,Palatino"><b>
597 <a name="simplechanges">Making simple changes</a>
598 </b></font></td></tr></table><ul>
600 There are some primitive transformation operations present in the LLVM
601 infrastructure that are worth knowing about. When performing
602 transformations, it's fairly common to manipulate the contents of
603 basic blocks. This section describes some of the common methods for
604 doing so and gives example code.
606 <!--_______________________________________________________________________-->
607 </ul><h4><a name="schanges_creating"><hr size=0>Creating and inserting
608 new <tt>Instruction</tt>s</h4><ul>
610 <i>Instantiating Instructions</i>
612 <p>Creation of <tt>Instruction</tt>s is straightforward: simply call the
613 constructor for the kind of instruction to instantiate and provide the
614 necessary parameters. For example, an <tt>AllocaInst</tt> only
615 <i>requires</i> a (const-ptr-to) <tt>Type</tt>. Thus:
617 <pre>AllocaInst* ai = new AllocaInst(Type::IntTy);</pre>
619 will create an <tt>AllocaInst</tt> instance that represents the
620 allocation of one integer in the current stack frame, at runtime.
621 Each <tt>Instruction</tt> subclass is likely to have varying default
622 parameters which change the semantics of the instruction, so refer to
623 the <a href="/doxygen/classInstruction.html">doxygen documentation for
624 the subclass of Instruction</a> that you're interested in
627 <p><i>Naming values</i></p>
630 It is very useful to name the values of instructions when you're able
631 to, as this facilitates the debugging of your transformations. If you
632 end up looking at generated LLVM machine code, you definitely want to
633 have logical names associated with the results of instructions! By
634 supplying a value for the <tt>Name</tt> (default) parameter of the
635 <tt>Instruction</tt> constructor, you associate a logical name with
636 the result of the instruction's execution at runtime. For example,
637 say that I'm writing a transformation that dynamically allocates space
638 for an integer on the stack, and that integer is going to be used as
639 some kind of index by some other code. To accomplish this, I place an
640 <tt>AllocaInst</tt> at the first point in the first
641 <tt>BasicBlock</tt> of some <tt>Function</tt>, and I'm intending to
642 use it within the same <tt>Function</tt>. I might do:
644 <pre>AllocaInst* pa = new AllocaInst(Type::IntTy, 0, "indexLoc");</pre>
646 where <tt>indexLoc</tt> is now the logical name of the instruction's
647 execution value, which is a pointer to an integer on the runtime
651 <p><i>Inserting instructions</i></p>
654 There are essentially two ways to insert an <tt>Instruction</tt> into
655 an existing sequence of instructions that form a <tt>BasicBlock</tt>:
657 <li>Insertion into an explicit instruction list
659 <p>Given a <tt>BasicBlock* pb</tt>, an <tt>Instruction* pi</tt> within
660 that <tt>BasicBlock</tt>, and a newly-created instruction
661 we wish to insert before <tt>*pi</tt>, we do the following:
664 BasicBlock* pb = ...;
665 Instruction* pi = ...;
666 Instruction* newInst = new Instruction(...);
667 pb->getInstList().insert(pi, newInst); // inserts newInst before pi in pb
671 <li>Insertion into an implicit instruction list
673 <tt>Instruction</tt> instances that are already in
674 <tt>BasicBlock</tt>s are implicitly associated with an existing
675 instruction list: the instruction list of the enclosing basic block.
676 Thus, we could have accomplished the same thing as the above code
677 without being given a <tt>BasicBlock</tt> by doing:
679 Instruction* pi = ...;
680 Instruction* newInst = new Instruction(...);
681 pi->getParent()->getInstList().insert(pi, newInst);
683 In fact, this sequence of steps occurs so frequently that the
684 <tt>Instruction</tt> class and <tt>Instruction</tt>-derived classes
685 provide constructors which take (as a default parameter) a pointer to
686 an <tt>Instruction</tt> which the newly-created <tt>Instruction</tt>
687 should precede. That is, <tt>Instruction</tt> constructors are
688 capable of inserting the newly-created instance into the
689 <tt>BasicBlock</tt> of a provided instruction, immediately before that
690 instruction. Using an <tt>Instruction</tt> constructor with a
691 <tt>insertBefore</tt> (default) parameter, the above code becomes:
693 Instruction* pi = ...;
694 Instruction* newInst = new Instruction(..., pi);
696 which is much cleaner, especially if you're creating a lot of
697 instructions and adding them to <tt>BasicBlock</tt>s.
702 <!--_______________________________________________________________________-->
703 </ul><h4><a name="schanges_deleting"><hr size=0>Deleting
704 <tt>Instruction</tt>s</h4><ul>
706 Deleting an instruction from an existing sequence of instructions that form a <a
707 href="#BasicBlock"><tt>BasicBlock</tt></a> is very straightforward. First, you
708 must have a pointer to the instruction that you wish to delete. Second, you
709 need to obtain the pointer to that instruction's basic block. You use the
710 pointer to the basic block to get its list of instructions and then use the
711 erase function to remove your instruction.<p>
716 <a href="#Instruction">Instruction</a> *I = .. ;
717 <a href="#BasicBlock">BasicBlock</a> *BB = I->getParent();
718 BB->getInstList().erase(I);
722 <!--_______________________________________________________________________-->
723 </ul><h4><a name="schanges_replacing"><hr size=0>Replacing an
724 <tt>Instruction</tt> with another <tt>Value</tt></h4><ul>
726 <!-- Value::replaceAllUsesWith
727 User::replaceUsesOfWith
728 Point out: include/llvm/Transforms/Utils/
729 especially BasicBlockUtils.h with:
730 ReplaceInstWithValue, ReplaceInstWithInst
734 <!-- *********************************************************************** -->
735 </ul><table width="100%" bgcolor="#330077" border=0 cellpadding=4 cellspacing=0>
736 <tr><td align=center><font color="#EEEEFF" size=+2 face="Georgia,Palatino"><b>
737 <a name="coreclasses">The Core LLVM Class Hierarchy Reference
738 </b></font></td></tr></table><ul>
739 <!-- *********************************************************************** -->
741 The Core LLVM classes are the primary means of representing the program being
742 inspected or transformed. The core LLVM classes are defined in header files in
743 the <tt>include/llvm/</tt> directory, and implemented in the <tt>lib/VMCore</tt>
747 <!-- ======================================================================= -->
748 </ul><table width="100%" bgcolor="#441188" border=0 cellpadding=4 cellspacing=0>
749 <tr><td> </td><td width="100%">
750 <font color="#EEEEFF" face="Georgia,Palatino"><b>
751 <a name="Value">The <tt>Value</tt> class</a>
752 </b></font></td></tr></table><ul>
754 <tt>#include "<a href="/doxygen/Value_8h-source.html">llvm/Value.h</a>"</tt></b><br>
755 doxygen info: <a href="/doxygen/classValue.html">Value Class</a><p>
758 The <tt>Value</tt> class is the most important class in LLVM Source base. It
759 represents a typed value that may be used (among other things) as an operand to
760 an instruction. There are many different types of <tt>Value</tt>s, such as <a
761 href="#Constant"><tt>Constant</tt></a>s, <a
762 href="#Argument"><tt>Argument</tt></a>s, and even <a
763 href="#Instruction"><tt>Instruction</tt></a>s and <a
764 href="#Function"><tt>Function</tt></a>s are <tt>Value</tt>s.<p>
766 A particular <tt>Value</tt> may be used many times in the LLVM representation
767 for a program. For example, an incoming argument to a function (represented
768 with an instance of the <a href="#Argument">Argument</a> class) is "used" by
769 every instruction in the function that references the argument. To keep track
770 of this relationship, the <tt>Value</tt> class keeps a list of all of the <a
771 href="#User"><tt>User</tt></a>s that is using it (the <a
772 href="#User"><tt>User</tt></a> class is a base class for all nodes in the LLVM
773 graph that can refer to <tt>Value</tt>s). This use list is how LLVM represents
774 def-use information in the program, and is accessible through the <tt>use_</tt>*
775 methods, shown below.<p>
777 Because LLVM is a typed representation, every LLVM <tt>Value</tt> is typed, and
778 this <a href="#Type">Type</a> is available through the <tt>getType()</tt>
779 method. <a name="#nameWarning">In addition, all LLVM values can be named. The
780 "name" of the <tt>Value</tt> is symbolic string printed in the LLVM code:<p>
783 %<b>foo</b> = add int 1, 2
786 The name of this instruction is "foo". <b>NOTE</b> that the name of any value
787 may be missing (an empty string), so names should <b>ONLY</b> be used for
788 debugging (making the source code easier to read, debugging printouts), they
789 should not be used to keep track of values or map between them. For this
790 purpose, use a <tt>std::map</tt> of pointers to the <tt>Value</tt> itself
793 One important aspect of LLVM is that there is no distinction between an SSA
794 variable and the operation that produces it. Because of this, any reference to
795 the value produced by an instruction (or the value available as an incoming
796 argument, for example) is represented as a direct pointer to the class that
797 represents this value. Although this may take some getting used to, it
798 simplifies the representation and makes it easier to manipulate.<p>
801 <!-- _______________________________________________________________________ -->
802 </ul><h4><a name="m_Value"><hr size=0>Important Public Members of
803 the <tt>Value</tt> class</h4><ul>
805 <li><tt>Value::use_iterator</tt> - Typedef for iterator over the use-list<br>
806 <tt>Value::use_const_iterator</tt>
807 - Typedef for const_iterator over the use-list<br>
808 <tt>unsigned use_size()</tt> - Returns the number of users of the value.<br>
809 <tt>bool use_empty()</tt> - Returns true if there are no users.<br>
810 <tt>use_iterator use_begin()</tt>
811 - Get an iterator to the start of the use-list.<br>
812 <tt>use_iterator use_end()</tt>
813 - Get an iterator to the end of the use-list.<br>
814 <tt><a href="#User">User</a> *use_back()</tt>
815 - Returns the last element in the list.<p>
817 These methods are the interface to access the def-use information in LLVM. As with all other iterators in LLVM, the naming conventions follow the conventions defined by the <a href="#stl">STL</a>.<p>
819 <li><tt><a href="#Type">Type</a> *getType() const</tt><p>
820 This method returns the Type of the Value.
822 <li><tt>bool hasName() const</tt><br>
823 <tt>std::string getName() const</tt><br>
824 <tt>void setName(const std::string &Name)</tt><p>
826 This family of methods is used to access and assign a name to a <tt>Value</tt>,
827 be aware of the <a href="#nameWarning">precaution above</a>.<p>
830 <li><tt>void replaceAllUsesWith(Value *V)</tt><p>
832 This method traverses the use list of a <tt>Value</tt> changing all <a
833 href="#User"><tt>User</tt>'s</a> of the current value to refer to "<tt>V</tt>"
834 instead. For example, if you detect that an instruction always produces a
835 constant value (for example through constant folding), you can replace all uses
836 of the instruction with the constant like this:<p>
839 Inst->replaceAllUsesWith(ConstVal);
844 <!-- ======================================================================= -->
845 </ul><table width="100%" bgcolor="#441188" border=0 cellpadding=4 cellspacing=0>
846 <tr><td> </td><td width="100%">
847 <font color="#EEEEFF" face="Georgia,Palatino"><b>
848 <a name="User">The <tt>User</tt> class</a>
849 </b></font></td></tr></table><ul>
851 <tt>#include "<a href="/doxygen/User_8h-source.html">llvm/User.h</a>"</tt></b><br>
852 doxygen info: <a href="/doxygen/classUser.html">User Class</a><br>
853 Superclass: <a href="#Value"><tt>Value</tt></a><p>
856 The <tt>User</tt> class is the common base class of all LLVM nodes that may
857 refer to <a href="#Value"><tt>Value</tt></a>s. It exposes a list of "Operands"
858 that are all of the <a href="#Value"><tt>Value</tt></a>s that the User is
859 referring to. The <tt>User</tt> class itself is a subclass of
862 The operands of a <tt>User</tt> point directly to the LLVM <a
863 href="#Value"><tt>Value</tt></a> that it refers to. Because LLVM uses Static
864 Single Assignment (SSA) form, there can only be one definition referred to,
865 allowing this direct connection. This connection provides the use-def
866 information in LLVM.<p>
868 <!-- _______________________________________________________________________ -->
869 </ul><h4><a name="m_User"><hr size=0>Important Public Members of
870 the <tt>User</tt> class</h4><ul>
872 The <tt>User</tt> class exposes the operand list in two ways: through an index
873 access interface and through an iterator based interface.<p>
875 <li><tt>Value *getOperand(unsigned i)</tt><br>
876 <tt>unsigned getNumOperands()</tt><p>
878 These two methods expose the operands of the <tt>User</tt> in a convenient form
879 for direct access.<p>
881 <li><tt>User::op_iterator</tt> - Typedef for iterator over the operand list<br>
882 <tt>User::op_const_iterator</tt>
883 <tt>use_iterator op_begin()</tt>
884 - Get an iterator to the start of the operand list.<br>
885 <tt>use_iterator op_end()</tt>
886 - Get an iterator to the end of the operand list.<p>
888 Together, these methods make up the iterator based interface to the operands of
893 <!-- ======================================================================= -->
894 </ul><table width="100%" bgcolor="#441188" border=0 cellpadding=4 cellspacing=0>
895 <tr><td> </td><td width="100%">
896 <font color="#EEEEFF" face="Georgia,Palatino"><b>
897 <a name="Instruction">The <tt>Instruction</tt> class</a>
898 </b></font></td></tr></table><ul>
901 href="/doxygen/Instruction_8h-source.html">llvm/Instruction.h</a>"</tt></b><br>
902 doxygen info: <a href="/doxygen/classInstruction.html">Instruction Class</a><br>
903 Superclasses: <a href="#User"><tt>User</tt></a>, <a
904 href="#Value"><tt>Value</tt></a><p>
906 The <tt>Instruction</tt> class is the common base class for all LLVM
907 instructions. It provides only a few methods, but is a very commonly used
908 class. The primary data tracked by the <tt>Instruction</tt> class itself is the
909 opcode (instruction type) and the parent <a
910 href="#BasicBlock"><tt>BasicBlock</tt></a> the <tt>Instruction</tt> is embedded
911 into. To represent a specific type of instruction, one of many subclasses of
912 <tt>Instruction</tt> are used.<p>
914 Because the <tt>Instruction</tt> class subclasses the <a
915 href="#User"><tt>User</tt></a> class, its operands can be accessed in the same
916 way as for other <a href="#User"><tt>User</tt></a>s (with the
917 <tt>getOperand()</tt>/<tt>getNumOperands()</tt> and
918 <tt>op_begin()</tt>/<tt>op_end()</tt> methods).<p>
920 An important file for the <tt>Instruction</tt> class is the
921 <tt>llvm/Instruction.def</tt> file. This file contains some meta-data about the
922 various different types of instructions in LLVM. It describes the enum values
923 that are used as opcodes (for example <tt>Instruction::Add</tt> and
924 <tt>Instruction::SetLE</tt>), as well as the concrete sub-classes of
925 <tt>Instruction</tt> that implement the instruction (for example <tt><a
926 href="#BinaryOperator">BinaryOperator</a></tt> and <tt><a
927 href="#SetCondInst">SetCondInst</a></tt>). Unfortunately, the use of macros in
928 this file confused doxygen, so these enum values don't show up correctly in the
929 <a href="/doxygen/classInstruction.html">doxygen output</a>.<p>
932 <!-- _______________________________________________________________________ -->
933 </ul><h4><a name="m_Instruction"><hr size=0>Important Public Members of
934 the <tt>Instruction</tt> class</h4><ul>
936 <li><tt><a href="#BasicBlock">BasicBlock</a> *getParent()</tt><p>
938 Returns the <a href="#BasicBlock"><tt>BasicBlock</tt></a> that this
939 <tt>Instruction</tt> is embedded into.<p>
941 <li><tt>bool hasSideEffects()</tt><p>
943 Returns true if the instruction has side effects, i.e. it is a <tt>call</tt>,
944 <tt>free</tt>, <tt>invoke</tt>, or <tt>store</tt>.<p>
946 <li><tt>unsigned getOpcode()</tt><p>
948 Returns the opcode for the <tt>Instruction</tt>.<p>
950 <li><tt><a href="#Instruction">Instruction</a> *clone() const</tt><p>
952 Returns another instance of the specified instruction, identical in all ways to
953 the original except that the instruction has no parent (ie it's not embedded
954 into a <a href="#BasicBlock"><tt>BasicBlock</tt></a>), and it has no name.<p>
960 \subsection{Subclasses of Instruction :}
962 <li>BinaryOperator : This subclass of Instruction defines a general interface to the all the instructions involvong binary operators in LLVM.
964 <li><tt>bool swapOperands()</tt>: Exchange the two operands to this instruction. If the instruction cannot be reversed (i.e. if it's a Div), it returns true.
966 <li>TerminatorInst : This subclass of Instructions defines an interface for all instructions that can terminate a BasicBlock.
968 <li> <tt>unsigned getNumSuccessors()</tt>: Returns the number of successors for this terminator instruction.
969 <li><tt>BasicBlock *getSuccessor(unsigned i)</tt>: As the name suggests returns the ith successor BasicBlock.
970 <li><tt>void setSuccessor(unsigned i, BasicBlock *B)</tt>: sets BasicBlock B as the ith succesor to this terminator instruction.
973 <li>PHINode : This represents the PHI instructions in the SSA form.
975 <li><tt> unsigned getNumIncomingValues()</tt>: Returns the number of incoming edges to this PHI node.
976 <li><tt> Value *getIncomingValue(unsigned i)</tt>: Returns the ith incoming Value.
977 <li><tt>void setIncomingValue(unsigned i, Value *V)</tt>: Sets the ith incoming Value as V
978 <li><tt>BasicBlock *getIncomingBlock(unsigned i)</tt>: Returns the Basic Block corresponding to the ith incoming Value.
979 <li><tt> void addIncoming(Value *D, BasicBlock *BB)</tt>:
980 Add an incoming value to the end of the PHI list
981 <li><tt> int getBasicBlockIndex(const BasicBlock *BB) const</tt>:
982 Returns the first index of the specified basic block in the value list for this PHI. Returns -1 if no instance.
984 <li>CastInst : In LLVM all casts have to be done through explicit cast instructions. CastInst defines the interface to the cast instructions.
985 <li>CallInst : This defines an interface to the call instruction in LLVM. ARguments to the function are nothing but operands of the instruction.
987 <li>: <tt>Function *getCalledFunction()</tt>: Returns a handle to the function that is being called by this Function.
989 <li>LoadInst, StoreInst, GetElemPtrInst : These subclasses represent load, store and getelementptr instructions in LLVM.
991 <li><tt>Value * getPointerOperand ()</tt>: Returns the Pointer Operand which is typically the 0th operand.
993 <li>BranchInst : This is a subclass of TerminatorInst and defines the interface for conditional and unconditional branches in LLVM.
995 <li><tt>bool isConditional()</tt>: Returns true if the branch is a conditional branch else returns false
996 <li> <tt>Value *getCondition()</tt>: Returns the condition if it is a conditional branch else returns null.
997 <li> <tt>void setUnconditionalDest(BasicBlock *Dest)</tt>: Changes the current branch to an unconditional one targetting the specified block.
1005 <!-- ======================================================================= -->
1006 </ul><table width="100%" bgcolor="#441188" border=0 cellpadding=4 cellspacing=0>
1007 <tr><td> </td><td width="100%">
1008 <font color="#EEEEFF" face="Georgia,Palatino"><b>
1009 <a name="BasicBlock">The <tt>BasicBlock</tt> class</a>
1010 </b></font></td></tr></table><ul>
1013 href="/doxygen/BasicBlock_8h-source.html">llvm/BasicBlock.h</a>"</tt></b><br>
1014 doxygen info: <a href="/doxygen/classBasicBlock.html">BasicBlock Class</a><br>
1015 Superclass: <a href="#Value"><tt>Value</tt></a><p>
1018 This class represents a single entry multiple exit section of the code, commonly
1019 known as a basic block by the compiler community. The <tt>BasicBlock</tt> class
1020 maintains a list of <a href="#Instruction"><tt>Instruction</tt></a>s, which form
1021 the body of the block. Matching the language definition, the last element of
1022 this list of instructions is always a terminator instruction (a subclass of the
1023 <a href="#TerminatorInst"><tt>TerminatorInst</tt></a> class).<p>
1025 In addition to tracking the list of instructions that make up the block, the
1026 <tt>BasicBlock</tt> class also keeps track of the <a
1027 href="#Function"><tt>Function</tt></a> that it is embedded into.<p>
1029 Note that <tt>BasicBlock</tt>s themselves are <a
1030 href="#Value"><tt>Value</tt></a>s, because they are referenced by instructions
1031 like branches and can go in the switch tables. <tt>BasicBlock</tt>s have type
1035 <!-- _______________________________________________________________________ -->
1036 </ul><h4><a name="m_BasicBlock"><hr size=0>Important Public Members of
1037 the <tt>BasicBlock</tt> class</h4><ul>
1039 <li><tt>BasicBlock(const std::string &Name = "", <a
1040 href="#Function">Function</a> *Parent = 0)</tt><p>
1042 The <tt>BasicBlock</tt> constructor is used to create new basic blocks for
1043 insertion into a function. The constructor simply takes a name for the new
1044 block, and optionally a <a href="#Function"><tt>Function</tt></a> to insert it
1045 into. If the <tt>Parent</tt> parameter is specified, the new
1046 <tt>BasicBlock</tt> is automatically inserted at the end of the specified <a
1047 href="#Function"><tt>Function</tt></a>, if not specified, the BasicBlock must be
1048 manually inserted into the <a href="#Function"><tt>Function</tt></a>.<p>
1050 <li><tt>BasicBlock::iterator</tt> - Typedef for instruction list iterator<br>
1051 <tt>BasicBlock::const_iterator</tt> - Typedef for const_iterator.<br>
1052 <tt>begin()</tt>, <tt>end()</tt>, <tt>front()</tt>, <tt>back()</tt>,
1053 <tt>size()</tt>, <tt>empty()</tt>, <tt>rbegin()</tt>, <tt>rend()</tt><p>
1055 These methods and typedefs are forwarding functions that have the same semantics
1056 as the standard library methods of the same names. These methods expose the
1057 underlying instruction list of a basic block in a way that is easy to
1058 manipulate. To get the full complement of container operations (including
1059 operations to update the list), you must use the <tt>getInstList()</tt>
1062 <li><tt>BasicBlock::InstListType &getInstList()</tt><p>
1064 This method is used to get access to the underlying container that actually
1065 holds the Instructions. This method must be used when there isn't a forwarding
1066 function in the <tt>BasicBlock</tt> class for the operation that you would like
1067 to perform. Because there are no forwarding functions for "updating"
1068 operations, you need to use this if you want to update the contents of a
1069 <tt>BasicBlock</tt>.<p>
1071 <li><tt><A href="#Function">Function</a> *getParent()</tt><p>
1073 Returns a pointer to <a href="#Function"><tt>Function</tt></a> the block is
1074 embedded into, or a null pointer if it is homeless.<p>
1076 <li><tt><a href="#TerminatorInst">TerminatorInst</a> *getTerminator()</tt><p>
1078 Returns a pointer to the terminator instruction that appears at the end of the
1079 <tt>BasicBlock</tt>. If there is no terminator instruction, or if the last
1080 instruction in the block is not a terminator, then a null pointer is
1084 <!-- ======================================================================= -->
1085 </ul><table width="100%" bgcolor="#441188" border=0 cellpadding=4 cellspacing=0>
1086 <tr><td> </td><td width="100%">
1087 <font color="#EEEEFF" face="Georgia,Palatino"><b>
1088 <a name="GlobalValue">The <tt>GlobalValue</tt> class</a>
1089 </b></font></td></tr></table><ul>
1092 href="/doxygen/GlobalValue_8h-source.html">llvm/GlobalValue.h</a>"</tt></b><br>
1093 doxygen info: <a href="/doxygen/classGlobalValue.html">GlobalValue Class</a><br>
1094 Superclasses: <a href="#User"><tt>User</tt></a>, <a
1095 href="#Value"><tt>Value</tt></a><p>
1097 Global values (<A href="#GlobalVariable"><tt>GlobalVariable</tt></a>s or <a
1098 href="#Function"><tt>Function</tt></a>s) are the only LLVM values that are
1099 visible in the bodies of all <a href="#Function"><tt>Function</tt></a>s.
1100 Because they are visible at global scope, they are also subject to linking with
1101 other globals defined in different translation units. To control the linking
1102 process, <tt>GlobalValue</tt>s know their linkage rules. Specifically,
1103 <tt>GlobalValue</tt>s know whether they have internal or external linkage.<p>
1105 If a <tt>GlobalValue</tt> has internal linkage (equivalent to being
1106 <tt>static</tt> in C), it is not visible to code outside the current translation
1107 unit, and does not participate in linking. If it has external linkage, it is
1108 visible to external code, and does participate in linking. In addition to
1109 linkage information, <tt>GlobalValue</tt>s keep track of which <a
1110 href="#Module"><tt>Module</tt></a> they are currently part of.<p>
1112 Because <tt>GlobalValue</tt>s are memory objects, they are always referred to by
1113 their address. As such, the <a href="#Type"><tt>Type</tt></a> of a global is
1114 always a pointer to its contents. This is explained in the LLVM Language
1115 Reference Manual.<p>
1118 <!-- _______________________________________________________________________ -->
1119 </ul><h4><a name="m_GlobalValue"><hr size=0>Important Public Members of
1120 the <tt>GlobalValue</tt> class</h4><ul>
1122 <li><tt>bool hasInternalLinkage() const</tt><br>
1123 <tt>bool hasExternalLinkage() const</tt><br>
1124 <tt>void setInternalLinkage(bool HasInternalLinkage)</tt><p>
1126 These methods manipulate the linkage characteristics of the
1127 <tt>GlobalValue</tt>.<p>
1129 <li><tt><a href="#Module">Module</a> *getParent()</tt><p>
1131 This returns the <a href="#Module"><tt>Module</tt></a> that the GlobalValue is
1132 currently embedded into.<p>
1136 <!-- ======================================================================= -->
1137 </ul><table width="100%" bgcolor="#441188" border=0 cellpadding=4 cellspacing=0>
1138 <tr><td> </td><td width="100%">
1139 <font color="#EEEEFF" face="Georgia,Palatino"><b>
1140 <a name="Function">The <tt>Function</tt> class</a>
1141 </b></font></td></tr></table><ul>
1144 href="/doxygen/Function_8h-source.html">llvm/Function.h</a>"</tt></b><br>
1145 doxygen info: <a href="/doxygen/classFunction.html">Function Class</a><br>
1146 Superclasses: <a href="#GlobalValue"><tt>GlobalValue</tt></a>, <a
1147 href="#User"><tt>User</tt></a>, <a href="#Value"><tt>Value</tt></a><p>
1149 The <tt>Function</tt> class represents a single procedure in LLVM. It is
1150 actually one of the more complex classes in the LLVM heirarchy because it must
1151 keep track of a large amount of data. The <tt>Function</tt> class keeps track
1152 of a list of <a href="#BasicBlock"><tt>BasicBlock</tt></a>s, a list of formal <a
1153 href="#Argument"><tt>Argument</tt></a>s, and a <a
1154 href="#SymbolTable"><tt>SymbolTable</tt></a>.<p>
1156 The list of <a href="#BasicBlock"><tt>BasicBlock</tt></a>s is the most commonly
1157 used part of <tt>Function</tt> objects. The list imposes an implicit ordering
1158 of the blocks in the function, which indicate how the code will be layed out by
1159 the backend. Additionally, the first <a
1160 href="#BasicBlock"><tt>BasicBlock</tt></a> is the implicit entry node for the
1161 <tt>Function</tt>. It is not legal in LLVM explicitly branch to this initial
1162 block. There are no implicit exit nodes, and in fact there may be multiple exit
1163 nodes from a single <tt>Function</tt>. If the <a
1164 href="#BasicBlock"><tt>BasicBlock</tt></a> list is empty, this indicates that
1165 the <tt>Function</tt> is actually a function declaration: the actual body of the
1166 function hasn't been linked in yet.<p>
1168 In addition to a list of <a href="#BasicBlock"><tt>BasicBlock</tt></a>s, the
1169 <tt>Function</tt> class also keeps track of the list of formal <a
1170 href="#Argument"><tt>Argument</tt></a>s that the function receives. This
1171 container manages the lifetime of the <a href="#Argument"><tt>Argument</tt></a>
1172 nodes, just like the <a href="#BasicBlock"><tt>BasicBlock</tt></a> list does for
1173 the <a href="#BasicBlock"><tt>BasicBlock</tt></a>s.<p>
1175 The <a href="#SymbolTable"><tt>SymbolTable</tt></a> is a very rarely used LLVM
1176 feature that is only used when you have to look up a value by name. Aside from
1177 that, the <a href="#SymbolTable"><tt>SymbolTable</tt></a> is used internally to
1178 make sure that there are not conflicts between the names of <a
1179 href="#Instruction"><tt>Instruction</tt></a>s, <a
1180 href="#BasicBlock"><tt>BasicBlock</tt></a>s, or <a
1181 href="#Argument"><tt>Argument</tt></a>s in the function body.<p>
1184 <!-- _______________________________________________________________________ -->
1185 </ul><h4><a name="m_Function"><hr size=0>Important Public Members of
1186 the <tt>Function</tt> class</h4><ul>
1188 <li><tt>Function(const <a href="#FunctionType">FunctionType</a> *Ty, bool isInternal, const std::string &N = "")</tt><p>
1190 Constructor used when you need to create new <tt>Function</tt>s to add the the
1191 program. The constructor must specify the type of the function to create and
1192 whether or not it should start out with internal or external linkage.<p>
1194 <li><tt>bool isExternal()</tt><p>
1196 Return whether or not the <tt>Function</tt> has a body defined. If the function
1197 is "external", it does not have a body, and thus must be resolved by linking
1198 with a function defined in a different translation unit.<p>
1201 <li><tt>Function::iterator</tt> - Typedef for basic block list iterator<br>
1202 <tt>Function::const_iterator</tt> - Typedef for const_iterator.<br>
1203 <tt>begin()</tt>, <tt>end()</tt>, <tt>front()</tt>, <tt>back()</tt>,
1204 <tt>size()</tt>, <tt>empty()</tt>, <tt>rbegin()</tt>, <tt>rend()</tt><p>
1206 These are forwarding methods that make it easy to access the contents of a
1207 <tt>Function</tt> object's <a href="#BasicBlock"><tt>BasicBlock</tt></a>
1210 <li><tt>Function::BasicBlockListType &getBasicBlockList()</tt><p>
1212 Returns the list of <a href="#BasicBlock"><tt>BasicBlock</tt></a>s. This is
1213 neccesary to use when you need to update the list or perform a complex action
1214 that doesn't have a forwarding method.<p>
1217 <li><tt>Function::aiterator</tt> - Typedef for the argument list iterator<br>
1218 <tt>Function::const_aiterator</tt> - Typedef for const_iterator.<br>
1219 <tt>abegin()</tt>, <tt>aend()</tt>, <tt>afront()</tt>, <tt>aback()</tt>,
1220 <tt>asize()</tt>, <tt>aempty()</tt>, <tt>arbegin()</tt>, <tt>arend()</tt><p>
1222 These are forwarding methods that make it easy to access the contents of a
1223 <tt>Function</tt> object's <a href="#Argument"><tt>Argument</tt></a> list.<p>
1225 <li><tt>Function::ArgumentListType &getArgumentList()</tt><p>
1227 Returns the list of <a href="#Argument"><tt>Argument</tt></a>s. This is
1228 neccesary to use when you need to update the list or perform a complex action
1229 that doesn't have a forwarding method.<p>
1233 <li><tt><a href="#BasicBlock">BasicBlock</a> &getEntryNode()</tt><p>
1235 Returns the entry <a href="#BasicBlock"><tt>BasicBlock</tt></a> for the
1236 function. Because the entry block for the function is always the first block,
1237 this returns the first block of the <tt>Function</tt>.<p>
1239 <li><tt><a href="#Type">Type</a> *getReturnType()</tt><br>
1240 <tt><a href="#FunctionType">FunctionType</a> *getFunctionType()</tt><p>
1242 This traverses the <a href="#Type"><tt>Type</tt></a> of the <tt>Function</tt>
1243 and returns the return type of the function, or the <a
1244 href="#FunctionType"><tt>FunctionType</tt></a> of the actual function.<p>
1247 <li><tt>bool hasSymbolTable() const</tt><p>
1249 Return true if the <tt>Function</tt> has a symbol table allocated to it and if
1250 there is at least one entry in it.<p>
1252 <li><tt><a href="#SymbolTable">SymbolTable</a> *getSymbolTable()</tt><p>
1254 Return a pointer to the <a href="#SymbolTable"><tt>SymbolTable</tt></a> for this
1255 <tt>Function</tt> or a null pointer if one has not been allocated (because there
1256 are no named values in the function).<p>
1258 <li><tt><a href="#SymbolTable">SymbolTable</a> *getSymbolTableSure()</tt><p>
1260 Return a pointer to the <a href="#SymbolTable"><tt>SymbolTable</tt></a> for this
1261 <tt>Function</tt> or allocate a new <a
1262 href="#SymbolTable"><tt>SymbolTable</tt></a> if one is not already around. This
1263 should only be used when adding elements to the <a
1264 href="#SymbolTable"><tt>SymbolTable</tt></a>, so that empty symbol tables are
1265 not left laying around.<p>
1269 <!-- ======================================================================= -->
1270 </ul><table width="100%" bgcolor="#441188" border=0 cellpadding=4 cellspacing=0>
1271 <tr><td> </td><td width="100%">
1272 <font color="#EEEEFF" face="Georgia,Palatino"><b>
1273 <a name="GlobalVariable">The <tt>GlobalVariable</tt> class</a>
1274 </b></font></td></tr></table><ul>
1277 href="/doxygen/GlobalVariable_8h-source.html">llvm/GlobalVariable.h</a>"</tt></b><br>
1278 doxygen info: <a href="/doxygen/classGlobalVariable.html">GlobalVariable Class</a><br>
1279 Superclasses: <a href="#GlobalValue"><tt>GlobalValue</tt></a>, <a
1280 href="#User"><tt>User</tt></a>, <a href="#Value"><tt>Value</tt></a><p>
1282 Global variables are represented with the (suprise suprise)
1283 <tt>GlobalVariable</tt> class. Like functions, <tt>GlobalVariable</tt>s are
1284 also subclasses of <a href="#GlobalValue"><tt>GlobalValue</tt></a>, and as such
1285 are always referenced by their address (global values must live in memory, so
1286 their "name" refers to their address). Global variables may have an initial
1287 value (which must be a <a href="#Constant"><tt>Constant</tt></a>), and if they
1288 have an initializer, they may be marked as "constant" themselves (indicating
1289 that their contents never change at runtime).<p>
1292 <!-- _______________________________________________________________________ -->
1293 </ul><h4><a name="m_GlobalVariable"><hr size=0>Important Public Members of the
1294 <tt>GlobalVariable</tt> class</h4><ul>
1296 <li><tt>GlobalVariable(const <a href="#Type">Type</a> *Ty, bool isConstant, bool
1297 isInternal, <a href="#Constant">Constant</a> *Initializer = 0, const std::string
1298 &Name = "")</tt><p>
1300 Create a new global variable of the specified type. If <tt>isConstant</tt> is
1301 true then the global variable will be marked as unchanging for the program, and
1302 if <tt>isInternal</tt> is true the resultant global variable will have internal
1303 linkage. Optionally an initializer and name may be specified for the global variable as well.<p>
1306 <li><tt>bool isConstant() const</tt><p>
1308 Returns true if this is a global variable is known not to be modified at
1312 <li><tt>bool hasInitializer()</tt><p>
1314 Returns true if this <tt>GlobalVariable</tt> has an intializer.<p>
1317 <li><tt><a href="#Constant">Constant</a> *getInitializer()</tt><p>
1319 Returns the intial value for a <tt>GlobalVariable</tt>. It is not legal to call
1320 this method if there is no initializer.<p>
1323 <!-- ======================================================================= -->
1324 </ul><table width="100%" bgcolor="#441188" border=0 cellpadding=4 cellspacing=0>
1325 <tr><td> </td><td width="100%">
1326 <font color="#EEEEFF" face="Georgia,Palatino"><b>
1327 <a name="Module">The <tt>Module</tt> class</a>
1328 </b></font></td></tr></table><ul>
1331 href="/doxygen/Module_8h-source.html">llvm/Module.h</a>"</tt></b><br>
1332 doxygen info: <a href="/doxygen/classModule.html">Module Class</a><p>
1334 The <tt>Module</tt> class represents the top level structure present in LLVM
1335 programs. An LLVM module is effectively either a translation unit of the
1336 original program or a combination of several translation units merged by the
1337 linker. The <tt>Module</tt> class keeps track of a list of <a
1338 href="#Function"><tt>Function</tt></a>s, a list of <a
1339 href="#GlobalVariable"><tt>GlobalVariable</tt></a>s, and a <a
1340 href="#SymbolTable"><tt>SymbolTable</tt></a>. Additionally, it contains a few
1341 helpful member functions that try to make common operations easy.<p>
1344 <!-- _______________________________________________________________________ -->
1345 </ul><h4><a name="m_Module"><hr size=0>Important Public Members of the
1346 <tt>Module</tt> class</h4><ul>
1348 <li><tt>Module::iterator</tt> - Typedef for function list iterator<br>
1349 <tt>Module::const_iterator</tt> - Typedef for const_iterator.<br>
1350 <tt>begin()</tt>, <tt>end()</tt>, <tt>front()</tt>, <tt>back()</tt>,
1351 <tt>size()</tt>, <tt>empty()</tt>, <tt>rbegin()</tt>, <tt>rend()</tt><p>
1353 These are forwarding methods that make it easy to access the contents of a
1354 <tt>Module</tt> object's <a href="#Function"><tt>Function</tt></a>
1357 <li><tt>Module::FunctionListType &getFunctionList()</tt><p>
1359 Returns the list of <a href="#Function"><tt>Function</tt></a>s. This is
1360 neccesary to use when you need to update the list or perform a complex action
1361 that doesn't have a forwarding method.<p>
1363 <!-- Global Variable -->
1366 <li><tt>Module::giterator</tt> - Typedef for global variable list iterator<br>
1367 <tt>Module::const_giterator</tt> - Typedef for const_iterator.<br>
1368 <tt>gbegin()</tt>, <tt>gend()</tt>, <tt>gfront()</tt>, <tt>gback()</tt>,
1369 <tt>gsize()</tt>, <tt>gempty()</tt>, <tt>grbegin()</tt>, <tt>grend()</tt><p>
1371 These are forwarding methods that make it easy to access the contents of a
1372 <tt>Module</tt> object's <a href="#GlobalVariable"><tt>GlobalVariable</tt></a>
1375 <li><tt>Module::GlobalListType &getGlobalList()</tt><p>
1377 Returns the list of <a href="#GlobalVariable"><tt>GlobalVariable</tt></a>s.
1378 This is neccesary to use when you need to update the list or perform a complex
1379 action that doesn't have a forwarding method.<p>
1382 <!-- Symbol table stuff -->
1385 <li><tt>bool hasSymbolTable() const</tt><p>
1387 Return true if the <tt>Module</tt> has a symbol table allocated to it and if
1388 there is at least one entry in it.<p>
1390 <li><tt><a href="#SymbolTable">SymbolTable</a> *getSymbolTable()</tt><p>
1392 Return a pointer to the <a href="#SymbolTable"><tt>SymbolTable</tt></a> for this
1393 <tt>Module</tt> or a null pointer if one has not been allocated (because there
1394 are no named values in the function).<p>
1396 <li><tt><a href="#SymbolTable">SymbolTable</a> *getSymbolTableSure()</tt><p>
1398 Return a pointer to the <a href="#SymbolTable"><tt>SymbolTable</tt></a> for this
1399 <tt>Module</tt> or allocate a new <a
1400 href="#SymbolTable"><tt>SymbolTable</tt></a> if one is not already around. This
1401 should only be used when adding elements to the <a
1402 href="#SymbolTable"><tt>SymbolTable</tt></a>, so that empty symbol tables are
1403 not left laying around.<p>
1406 <!-- Convenience methods -->
1409 <li><tt><a href="#Function">Function</a> *getFunction(const std::string &Name, const <a href="#FunctionType">FunctionType</a> *Ty)</tt><p>
1411 Look up the specified function in the <tt>Module</tt> <a
1412 href="#SymbolTable"><tt>SymbolTable</tt></a>. If it does not exist, return
1416 <li><tt><a href="#Function">Function</a> *getOrInsertFunction(const std::string
1417 &Name, const <a href="#FunctionType">FunctionType</a> *T)</tt><p>
1419 Look up the specified function in the <tt>Module</tt> <a
1420 href="#SymbolTable"><tt>SymbolTable</tt></a>. If it does not exist, add an
1421 external declaration for the function and return it.<p>
1424 <li><tt>std::string getTypeName(const <a href="#Type">Type</a> *Ty)</tt><p>
1426 If there is at least one entry in the <a
1427 href="#SymbolTable"><tt>SymbolTable</tt></a> for the specified <a
1428 href="#Type"><tt>Type</tt></a>, return it. Otherwise return the empty
1432 <li><tt>bool addTypeName(const std::string &Name, const <a href="#Type">Type</a>
1435 Insert an entry in the <a href="#SymbolTable"><tt>SymbolTable</tt></a> mapping
1436 <tt>Name</tt> to <tt>Ty</tt>. If there is already an entry for this name, true
1437 is returned and the <a href="#SymbolTable"><tt>SymbolTable</tt></a> is not
1441 <!-- ======================================================================= -->
1442 </ul><table width="100%" bgcolor="#441188" border=0 cellpadding=4 cellspacing=0>
1443 <tr><td> </td><td width="100%">
1444 <font color="#EEEEFF" face="Georgia,Palatino"><b>
1445 <a name="Constant">The <tt>Constant</tt> class and subclasses</a>
1446 </b></font></td></tr></table><ul>
1448 Constant represents a base class for different types of constants. It is
1449 subclassed by ConstantBool, ConstantInt, ConstantSInt, ConstantUInt,
1450 ConstantArray etc for representing the various types of Constants.<p>
1453 <!-- _______________________________________________________________________ -->
1454 </ul><h4><a name="m_Value"><hr size=0>Important Public Methods</h4><ul>
1456 <li><tt>bool isConstantExpr()</tt>: Returns true if it is a ConstantExpr
1461 \subsection{Important Subclasses of Constant}
1463 <li>ConstantSInt : This subclass of Constant represents a signed integer constant.
1465 <li><tt>int64_t getValue () const</tt>: Returns the underlying value of this constant.
1467 <li>ConstantUInt : This class represents an unsigned integer.
1469 <li><tt>uint64_t getValue () const</tt>: Returns the underlying value of this constant.
1471 <li>ConstantFP : This class represents a floating point constant.
1473 <li><tt>double getValue () const</tt>: Returns the underlying value of this constant.
1475 <li>ConstantBool : This represents a boolean constant.
1477 <li><tt>bool getValue () const</tt>: Returns the underlying value of this constant.
1479 <li>ConstantArray : This represents a constant array.
1481 <li><tt>const std::vector<Use> &getValues() const</tt>: Returns a Vecotr of component constants that makeup this array.
1483 <li>ConstantStruct : This represents a constant struct.
1485 <li><tt>const std::vector<Use> &getValues() const</tt>: Returns a Vecotr of component constants that makeup this array.
1487 <li>ConstantPointerRef : This represents a constant pointer value that is initialized to point to a global value, which lies at a constant fixed address.
1489 <li><tt>GlobalValue *getValue()</tt>: Returns the global value to which this pointer is pointing to.
1494 <!-- ======================================================================= -->
1495 </ul><table width="100%" bgcolor="#441188" border=0 cellpadding=4 cellspacing=0>
1496 <tr><td> </td><td width="100%">
1497 <font color="#EEEEFF" face="Georgia,Palatino"><b>
1498 <a name="Type">The <tt>Type</tt> class and Derived Types</a>
1499 </b></font></td></tr></table><ul>
1501 Type as noted earlier is also a subclass of a Value class. Any primitive
1502 type (like int, short etc) in LLVM is an instance of Type Class. All
1503 other types are instances of subclasses of type like FunctionType,
1504 ArrayType etc. DerivedType is the interface for all such dervied types
1505 including FunctionType, ArrayType, PointerType, StructType. Types can have
1506 names. They can be recursive (StructType). There exists exactly one instance
1507 of any type structure at a time. This allows using pointer equality of Type *s for comparing types.
1509 <!-- _______________________________________________________________________ -->
1510 </ul><h4><a name="m_Value"><hr size=0>Important Public Methods</h4><ul>
1512 <li><tt>PrimitiveID getPrimitiveID () const</tt>: Returns the base type of the type.
1513 <li><tt> bool isSigned () const</tt>: Returns whether an integral numeric type is signed. This is true for SByteTy, ShortTy, IntTy, LongTy. Note that this is not true for Float and Double.
1514 <li><tt>bool isUnsigned () const</tt>: Returns whether a numeric type is unsigned. This is not quite the complement of isSigned... nonnumeric types return false as they do with isSigned. This returns true for UByteTy, UShortTy, UIntTy, and ULongTy.
1515 <li><tt> bool isInteger () const</tt>: Equilivent to isSigned() || isUnsigned(), but with only a single virtual function invocation.
1516 <li><tt>bool isIntegral () const</tt>: Returns true if this is an integral type, which is either Bool type or one of the Integer types.
1518 <li><tt>bool isFloatingPoint ()</tt>: Return true if this is one of the two floating point types.
1519 <li><tt>bool isRecursive () const</tt>: Returns rue if the type graph contains a cycle.
1520 <li><tt>isLosslesslyConvertableTo (const Type *Ty) const</tt>: Return true if this type can be converted to 'Ty' without any reinterpretation of bits. For example, uint to int.
1521 <li><tt>bool isPrimitiveType () const</tt>: Returns true if it is a primitive type.
1522 <li><tt>bool isDerivedType () const</tt>: Returns true if it is a derived type.
1523 <li><tt>const Type * getContainedType (unsigned i) const</tt>:
1524 This method is used to implement the type iterator. For derived types, this returns the types 'contained' in the derived type, returning 0 when 'i' becomes invalid. This allows the user to iterate over the types in a struct, for example, really easily.
1525 <li><tt>unsigned getNumContainedTypes () const</tt>: Return the number of types in the derived type.
1529 \subsection{Derived Types}
1531 <li>SequentialType : This is subclassed by ArrayType and PointerType
1533 <li><tt>const Type * getElementType () const</tt>: Returns the type of each of the elements in the sequential type.
1535 <li>ArrayType : This is a subclass of SequentialType and defines interface for array types.
1537 <li><tt>unsigned getNumElements () const</tt>: Returns the number of elements in the array.
1539 <li>PointerType : Subclass of SequentialType for pointer types.
1540 <li>StructType : subclass of DerivedTypes for struct types
1541 <li>FunctionType : subclass of DerivedTypes for function types.
1544 <li><tt>bool isVarArg () const</tt>: Returns true if its a vararg function
1545 <li><tt> const Type * getReturnType () const</tt>: Returns the return type of the function.
1546 <li><tt> const ParamTypes &getParamTypes () const</tt>: Returns a vector of parameter types.
1547 <li><tt>const Type * getParamType (unsigned i)</tt>: Returns the type of the ith parameter.
1548 <li><tt> const unsigned getNumParams () const</tt>: Returns the number of formal parameters.
1555 <!-- ======================================================================= -->
1556 </ul><table width="100%" bgcolor="#441188" border=0 cellpadding=4 cellspacing=0>
1557 <tr><td> </td><td width="100%">
1558 <font color="#EEEEFF" face="Georgia,Palatino"><b>
1559 <a name="Argument">The <tt>Argument</tt> class</a>
1560 </b></font></td></tr></table><ul>
1562 This subclass of Value defines the interface for incoming formal arguments to a
1563 function. A Function maitanis a list of its formal arguments. An argument has a
1564 pointer to the parent Function.
1569 <!-- *********************************************************************** -->
1571 <!-- *********************************************************************** -->
1574 <address>By: <a href="mailto:dhurjati@cs.uiuc.edu">Dinakar Dhurjati</a> and
1575 <a href="mailto:sabre@nondot.org">Chris Lattner</a></address>
1576 <!-- Created: Tue Aug 6 15:00:33 CDT 2002 -->
1577 <!-- hhmts start -->
1578 Last modified: Thu Sep 12 14:06:40 CDT 2002
1580 </font></body></html>