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>
142 <li><a href="#UserLayout">The <tt>User</tt> and owned <tt>Use</tt> classes' memory layout</a></li>
145 <li><a href="#coreclasses">The Core LLVM Class Hierarchy Reference</a>
147 <li><a href="#Type">The <tt>Type</tt> class</a> </li>
148 <li><a href="#Module">The <tt>Module</tt> class</a></li>
149 <li><a href="#Value">The <tt>Value</tt> class</a>
151 <li><a href="#User">The <tt>User</tt> class</a>
153 <li><a href="#Instruction">The <tt>Instruction</tt> class</a></li>
154 <li><a href="#Constant">The <tt>Constant</tt> class</a>
156 <li><a href="#GlobalValue">The <tt>GlobalValue</tt> class</a>
158 <li><a href="#Function">The <tt>Function</tt> class</a></li>
159 <li><a href="#GlobalVariable">The <tt>GlobalVariable</tt> class</a></li>
166 <li><a href="#BasicBlock">The <tt>BasicBlock</tt> class</a></li>
167 <li><a href="#Argument">The <tt>Argument</tt> class</a></li>
174 <div class="doc_author">
175 <p>Written by <a href="mailto:sabre@nondot.org">Chris Lattner</a>,
176 <a href="mailto:dhurjati@cs.uiuc.edu">Dinakar Dhurjati</a>,
177 <a href="mailto:ggreif@gmail.com">Gabor Greif</a>,
178 <a href="mailto:jstanley@cs.uiuc.edu">Joel Stanley</a> and
179 <a href="mailto:rspencer@x10sys.com">Reid Spencer</a></p>
182 <!-- *********************************************************************** -->
183 <div class="doc_section">
184 <a name="introduction">Introduction </a>
186 <!-- *********************************************************************** -->
188 <div class="doc_text">
190 <p>This document is meant to highlight some of the important classes and
191 interfaces available in the LLVM source-base. This manual is not
192 intended to explain what LLVM is, how it works, and what LLVM code looks
193 like. It assumes that you know the basics of LLVM and are interested
194 in writing transformations or otherwise analyzing or manipulating the
197 <p>This document should get you oriented so that you can find your
198 way in the continuously growing source code that makes up the LLVM
199 infrastructure. Note that this manual is not intended to serve as a
200 replacement for reading the source code, so if you think there should be
201 a method in one of these classes to do something, but it's not listed,
202 check the source. Links to the <a href="/doxygen/">doxygen</a> sources
203 are provided to make this as easy as possible.</p>
205 <p>The first section of this document describes general information that is
206 useful to know when working in the LLVM infrastructure, and the second describes
207 the Core LLVM classes. In the future this manual will be extended with
208 information describing how to use extension libraries, such as dominator
209 information, CFG traversal routines, and useful utilities like the <tt><a
210 href="/doxygen/InstVisitor_8h-source.html">InstVisitor</a></tt> template.</p>
214 <!-- *********************************************************************** -->
215 <div class="doc_section">
216 <a name="general">General Information</a>
218 <!-- *********************************************************************** -->
220 <div class="doc_text">
222 <p>This section contains general information that is useful if you are working
223 in the LLVM source-base, but that isn't specific to any particular API.</p>
227 <!-- ======================================================================= -->
228 <div class="doc_subsection">
229 <a name="stl">The C++ Standard Template Library</a>
232 <div class="doc_text">
234 <p>LLVM makes heavy use of the C++ Standard Template Library (STL),
235 perhaps much more than you are used to, or have seen before. Because of
236 this, you might want to do a little background reading in the
237 techniques used and capabilities of the library. There are many good
238 pages that discuss the STL, and several books on the subject that you
239 can get, so it will not be discussed in this document.</p>
241 <p>Here are some useful links:</p>
245 <li><a href="http://www.dinkumware.com/refxcpp.html">Dinkumware C++ Library
246 reference</a> - an excellent reference for the STL and other parts of the
247 standard C++ library.</li>
249 <li><a href="http://www.tempest-sw.com/cpp/">C++ In a Nutshell</a> - This is an
250 O'Reilly book in the making. It has a decent
252 Reference that rivals Dinkumware's, and is unfortunately no longer free since the book has been
255 <li><a href="http://www.parashift.com/c++-faq-lite/">C++ Frequently Asked
258 <li><a href="http://www.sgi.com/tech/stl/">SGI's STL Programmer's Guide</a> -
260 href="http://www.sgi.com/tech/stl/stl_introduction.html">Introduction to the
263 <li><a href="http://www.research.att.com/%7Ebs/C++.html">Bjarne Stroustrup's C++
266 <li><a href="http://64.78.49.204/">
267 Bruce Eckel's Thinking in C++, 2nd ed. Volume 2 Revision 4.0 (even better, get
272 <p>You are also encouraged to take a look at the <a
273 href="CodingStandards.html">LLVM Coding Standards</a> guide which focuses on how
274 to write maintainable code more than where to put your curly braces.</p>
278 <!-- ======================================================================= -->
279 <div class="doc_subsection">
280 <a name="stl">Other useful references</a>
283 <div class="doc_text">
286 <li><a href="http://www.psc.edu/%7Esemke/cvs_branches.html">CVS
287 Branch and Tag Primer</a></li>
288 <li><a href="http://www.fortran-2000.com/ArnaudRecipes/sharedlib.html">Using
289 static and shared libraries across platforms</a></li>
294 <!-- *********************************************************************** -->
295 <div class="doc_section">
296 <a name="apis">Important and useful LLVM APIs</a>
298 <!-- *********************************************************************** -->
300 <div class="doc_text">
302 <p>Here we highlight some LLVM APIs that are generally useful and good to
303 know about when writing transformations.</p>
307 <!-- ======================================================================= -->
308 <div class="doc_subsection">
309 <a name="isa">The <tt>isa<></tt>, <tt>cast<></tt> and
310 <tt>dyn_cast<></tt> templates</a>
313 <div class="doc_text">
315 <p>The LLVM source-base makes extensive use of a custom form of RTTI.
316 These templates have many similarities to the C++ <tt>dynamic_cast<></tt>
317 operator, but they don't have some drawbacks (primarily stemming from
318 the fact that <tt>dynamic_cast<></tt> only works on classes that
319 have a v-table). Because they are used so often, you must know what they
320 do and how they work. All of these templates are defined in the <a
321 href="/doxygen/Casting_8h-source.html"><tt>llvm/Support/Casting.h</tt></a>
322 file (note that you very rarely have to include this file directly).</p>
325 <dt><tt>isa<></tt>: </dt>
327 <dd><p>The <tt>isa<></tt> operator works exactly like the Java
328 "<tt>instanceof</tt>" operator. It returns true or false depending on whether
329 a reference or pointer points to an instance of the specified class. This can
330 be very useful for constraint checking of various sorts (example below).</p>
333 <dt><tt>cast<></tt>: </dt>
335 <dd><p>The <tt>cast<></tt> operator is a "checked cast" operation. It
336 converts a pointer or reference from a base class to a derived class, causing
337 an assertion failure if it is not really an instance of the right type. This
338 should be used in cases where you have some information that makes you believe
339 that something is of the right type. An example of the <tt>isa<></tt>
340 and <tt>cast<></tt> template is:</p>
342 <div class="doc_code">
344 static bool isLoopInvariant(const <a href="#Value">Value</a> *V, const Loop *L) {
345 if (isa<<a href="#Constant">Constant</a>>(V) || isa<<a href="#Argument">Argument</a>>(V) || isa<<a href="#GlobalValue">GlobalValue</a>>(V))
348 // <i>Otherwise, it must be an instruction...</i>
349 return !L->contains(cast<<a href="#Instruction">Instruction</a>>(V)->getParent());
354 <p>Note that you should <b>not</b> use an <tt>isa<></tt> test followed
355 by a <tt>cast<></tt>, for that use the <tt>dyn_cast<></tt>
360 <dt><tt>dyn_cast<></tt>:</dt>
362 <dd><p>The <tt>dyn_cast<></tt> operator is a "checking cast" operation.
363 It checks to see if the operand is of the specified type, and if so, returns a
364 pointer to it (this operator does not work with references). If the operand is
365 not of the correct type, a null pointer is returned. Thus, this works very
366 much like the <tt>dynamic_cast<></tt> operator in C++, and should be
367 used in the same circumstances. Typically, the <tt>dyn_cast<></tt>
368 operator is used in an <tt>if</tt> statement or some other flow control
369 statement like this:</p>
371 <div class="doc_code">
373 if (<a href="#AllocationInst">AllocationInst</a> *AI = dyn_cast<<a href="#AllocationInst">AllocationInst</a>>(Val)) {
379 <p>This form of the <tt>if</tt> statement effectively combines together a call
380 to <tt>isa<></tt> and a call to <tt>cast<></tt> into one
381 statement, which is very convenient.</p>
383 <p>Note that the <tt>dyn_cast<></tt> operator, like C++'s
384 <tt>dynamic_cast<></tt> or Java's <tt>instanceof</tt> operator, can be
385 abused. In particular, you should not use big chained <tt>if/then/else</tt>
386 blocks to check for lots of different variants of classes. If you find
387 yourself wanting to do this, it is much cleaner and more efficient to use the
388 <tt>InstVisitor</tt> class to dispatch over the instruction type directly.</p>
392 <dt><tt>cast_or_null<></tt>: </dt>
394 <dd><p>The <tt>cast_or_null<></tt> operator works just like the
395 <tt>cast<></tt> operator, except that it allows for a null pointer as an
396 argument (which it then propagates). This can sometimes be useful, allowing
397 you to combine several null checks into one.</p></dd>
399 <dt><tt>dyn_cast_or_null<></tt>: </dt>
401 <dd><p>The <tt>dyn_cast_or_null<></tt> operator works just like the
402 <tt>dyn_cast<></tt> operator, except that it allows for a null pointer
403 as an argument (which it then propagates). This can sometimes be useful,
404 allowing you to combine several null checks into one.</p></dd>
408 <p>These five templates can be used with any classes, whether they have a
409 v-table or not. To add support for these templates, you simply need to add
410 <tt>classof</tt> static methods to the class you are interested casting
411 to. Describing this is currently outside the scope of this document, but there
412 are lots of examples in the LLVM source base.</p>
416 <!-- ======================================================================= -->
417 <div class="doc_subsection">
418 <a name="DEBUG">The <tt>DEBUG()</tt> macro and <tt>-debug</tt> option</a>
421 <div class="doc_text">
423 <p>Often when working on your pass you will put a bunch of debugging printouts
424 and other code into your pass. After you get it working, you want to remove
425 it, but you may need it again in the future (to work out new bugs that you run
428 <p> Naturally, because of this, you don't want to delete the debug printouts,
429 but you don't want them to always be noisy. A standard compromise is to comment
430 them out, allowing you to enable them if you need them in the future.</p>
432 <p>The "<tt><a href="/doxygen/Debug_8h-source.html">llvm/Support/Debug.h</a></tt>"
433 file provides a macro named <tt>DEBUG()</tt> that is a much nicer solution to
434 this problem. Basically, you can put arbitrary code into the argument of the
435 <tt>DEBUG</tt> macro, and it is only executed if '<tt>opt</tt>' (or any other
436 tool) is run with the '<tt>-debug</tt>' command line argument:</p>
438 <div class="doc_code">
440 DOUT << "I am here!\n";
444 <p>Then you can run your pass like this:</p>
446 <div class="doc_code">
448 $ opt < a.bc > /dev/null -mypass
449 <i><no output></i>
450 $ opt < a.bc > /dev/null -mypass -debug
455 <p>Using the <tt>DEBUG()</tt> macro instead of a home-brewed solution allows you
456 to not have to create "yet another" command line option for the debug output for
457 your pass. Note that <tt>DEBUG()</tt> macros are disabled for optimized builds,
458 so they do not cause a performance impact at all (for the same reason, they
459 should also not contain side-effects!).</p>
461 <p>One additional nice thing about the <tt>DEBUG()</tt> macro is that you can
462 enable or disable it directly in gdb. Just use "<tt>set DebugFlag=0</tt>" or
463 "<tt>set DebugFlag=1</tt>" from the gdb if the program is running. If the
464 program hasn't been started yet, you can always just run it with
469 <!-- _______________________________________________________________________ -->
470 <div class="doc_subsubsection">
471 <a name="DEBUG_TYPE">Fine grained debug info with <tt>DEBUG_TYPE</tt> and
472 the <tt>-debug-only</tt> option</a>
475 <div class="doc_text">
477 <p>Sometimes you may find yourself in a situation where enabling <tt>-debug</tt>
478 just turns on <b>too much</b> information (such as when working on the code
479 generator). If you want to enable debug information with more fine-grained
480 control, you define the <tt>DEBUG_TYPE</tt> macro and the <tt>-debug</tt> only
481 option as follows:</p>
483 <div class="doc_code">
485 DOUT << "No debug type\n";
487 #define DEBUG_TYPE "foo"
488 DOUT << "'foo' debug type\n";
490 #define DEBUG_TYPE "bar"
491 DOUT << "'bar' debug type\n";
493 #define DEBUG_TYPE ""
494 DOUT << "No debug type (2)\n";
498 <p>Then you can run your pass like this:</p>
500 <div class="doc_code">
502 $ opt < a.bc > /dev/null -mypass
503 <i><no output></i>
504 $ opt < a.bc > /dev/null -mypass -debug
509 $ opt < a.bc > /dev/null -mypass -debug-only=foo
511 $ opt < a.bc > /dev/null -mypass -debug-only=bar
516 <p>Of course, in practice, you should only set <tt>DEBUG_TYPE</tt> at the top of
517 a file, to specify the debug type for the entire module (if you do this before
518 you <tt>#include "llvm/Support/Debug.h"</tt>, you don't have to insert the ugly
519 <tt>#undef</tt>'s). Also, you should use names more meaningful than "foo" and
520 "bar", because there is no system in place to ensure that names do not
521 conflict. If two different modules use the same string, they will all be turned
522 on when the name is specified. This allows, for example, all debug information
523 for instruction scheduling to be enabled with <tt>-debug-type=InstrSched</tt>,
524 even if the source lives in multiple files.</p>
528 <!-- ======================================================================= -->
529 <div class="doc_subsection">
530 <a name="Statistic">The <tt>Statistic</tt> class & <tt>-stats</tt>
534 <div class="doc_text">
537 href="/doxygen/Statistic_8h-source.html">llvm/ADT/Statistic.h</a></tt>" file
538 provides a class named <tt>Statistic</tt> that is used as a unified way to
539 keep track of what the LLVM compiler is doing and how effective various
540 optimizations are. It is useful to see what optimizations are contributing to
541 making a particular program run faster.</p>
543 <p>Often you may run your pass on some big program, and you're interested to see
544 how many times it makes a certain transformation. Although you can do this with
545 hand inspection, or some ad-hoc method, this is a real pain and not very useful
546 for big programs. Using the <tt>Statistic</tt> class makes it very easy to
547 keep track of this information, and the calculated information is presented in a
548 uniform manner with the rest of the passes being executed.</p>
550 <p>There are many examples of <tt>Statistic</tt> uses, but the basics of using
551 it are as follows:</p>
554 <li><p>Define your statistic like this:</p>
556 <div class="doc_code">
558 #define <a href="#DEBUG_TYPE">DEBUG_TYPE</a> "mypassname" <i>// This goes before any #includes.</i>
559 STATISTIC(NumXForms, "The # of times I did stuff");
563 <p>The <tt>STATISTIC</tt> macro defines a static variable, whose name is
564 specified by the first argument. The pass name is taken from the DEBUG_TYPE
565 macro, and the description is taken from the second argument. The variable
566 defined ("NumXForms" in this case) acts like an unsigned integer.</p></li>
568 <li><p>Whenever you make a transformation, bump the counter:</p>
570 <div class="doc_code">
572 ++NumXForms; // <i>I did stuff!</i>
579 <p>That's all you have to do. To get '<tt>opt</tt>' to print out the
580 statistics gathered, use the '<tt>-stats</tt>' option:</p>
582 <div class="doc_code">
584 $ opt -stats -mypassname < program.bc > /dev/null
585 <i>... statistics output ...</i>
589 <p> When running <tt>opt</tt> on a C file from the SPEC benchmark
590 suite, it gives a report that looks like this:</p>
592 <div class="doc_code">
594 7646 bitcodewriter - Number of normal instructions
595 725 bitcodewriter - Number of oversized instructions
596 129996 bitcodewriter - Number of bitcode bytes written
597 2817 raise - Number of insts DCEd or constprop'd
598 3213 raise - Number of cast-of-self removed
599 5046 raise - Number of expression trees converted
600 75 raise - Number of other getelementptr's formed
601 138 raise - Number of load/store peepholes
602 42 deadtypeelim - Number of unused typenames removed from symtab
603 392 funcresolve - Number of varargs functions resolved
604 27 globaldce - Number of global variables removed
605 2 adce - Number of basic blocks removed
606 134 cee - Number of branches revectored
607 49 cee - Number of setcc instruction eliminated
608 532 gcse - Number of loads removed
609 2919 gcse - Number of instructions removed
610 86 indvars - Number of canonical indvars added
611 87 indvars - Number of aux indvars removed
612 25 instcombine - Number of dead inst eliminate
613 434 instcombine - Number of insts combined
614 248 licm - Number of load insts hoisted
615 1298 licm - Number of insts hoisted to a loop pre-header
616 3 licm - Number of insts hoisted to multiple loop preds (bad, no loop pre-header)
617 75 mem2reg - Number of alloca's promoted
618 1444 cfgsimplify - Number of blocks simplified
622 <p>Obviously, with so many optimizations, having a unified framework for this
623 stuff is very nice. Making your pass fit well into the framework makes it more
624 maintainable and useful.</p>
628 <!-- ======================================================================= -->
629 <div class="doc_subsection">
630 <a name="ViewGraph">Viewing graphs while debugging code</a>
633 <div class="doc_text">
635 <p>Several of the important data structures in LLVM are graphs: for example
636 CFGs made out of LLVM <a href="#BasicBlock">BasicBlock</a>s, CFGs made out of
637 LLVM <a href="CodeGenerator.html#machinebasicblock">MachineBasicBlock</a>s, and
638 <a href="CodeGenerator.html#selectiondag_intro">Instruction Selection
639 DAGs</a>. In many cases, while debugging various parts of the compiler, it is
640 nice to instantly visualize these graphs.</p>
642 <p>LLVM provides several callbacks that are available in a debug build to do
643 exactly that. If you call the <tt>Function::viewCFG()</tt> method, for example,
644 the current LLVM tool will pop up a window containing the CFG for the function
645 where each basic block is a node in the graph, and each node contains the
646 instructions in the block. Similarly, there also exists
647 <tt>Function::viewCFGOnly()</tt> (does not include the instructions), the
648 <tt>MachineFunction::viewCFG()</tt> and <tt>MachineFunction::viewCFGOnly()</tt>,
649 and the <tt>SelectionDAG::viewGraph()</tt> methods. Within GDB, for example,
650 you can usually use something like <tt>call DAG.viewGraph()</tt> to pop
651 up a window. Alternatively, you can sprinkle calls to these functions in your
652 code in places you want to debug.</p>
654 <p>Getting this to work requires a small amount of configuration. On Unix
655 systems with X11, install the <a href="http://www.graphviz.org">graphviz</a>
656 toolkit, and make sure 'dot' and 'gv' are in your path. If you are running on
657 Mac OS/X, download and install the Mac OS/X <a
658 href="http://www.pixelglow.com/graphviz/">Graphviz program</a>, and add
659 <tt>/Applications/Graphviz.app/Contents/MacOS/</tt> (or wherever you install
660 it) to your path. Once in your system and path are set up, rerun the LLVM
661 configure script and rebuild LLVM to enable this functionality.</p>
663 <p><tt>SelectionDAG</tt> has been extended to make it easier to locate
664 <i>interesting</i> nodes in large complex graphs. From gdb, if you
665 <tt>call DAG.setGraphColor(<i>node</i>, "<i>color</i>")</tt>, then the
666 next <tt>call DAG.viewGraph()</tt> would highlight the node in the
667 specified color (choices of colors can be found at <a
668 href="http://www.graphviz.org/doc/info/colors.html">colors</a>.) More
669 complex node attributes can be provided with <tt>call
670 DAG.setGraphAttrs(<i>node</i>, "<i>attributes</i>")</tt> (choices can be
671 found at <a href="http://www.graphviz.org/doc/info/attrs.html">Graph
672 Attributes</a>.) If you want to restart and clear all the current graph
673 attributes, then you can <tt>call DAG.clearGraphAttrs()</tt>. </p>
677 <!-- *********************************************************************** -->
678 <div class="doc_section">
679 <a name="datastructure">Picking the Right Data Structure for a Task</a>
681 <!-- *********************************************************************** -->
683 <div class="doc_text">
685 <p>LLVM has a plethora of data structures in the <tt>llvm/ADT/</tt> directory,
686 and we commonly use STL data structures. This section describes the trade-offs
687 you should consider when you pick one.</p>
690 The first step is a choose your own adventure: do you want a sequential
691 container, a set-like container, or a map-like container? The most important
692 thing when choosing a container is the algorithmic properties of how you plan to
693 access the container. Based on that, you should use:</p>
696 <li>a <a href="#ds_map">map-like</a> container if you need efficient look-up
697 of an value based on another value. Map-like containers also support
698 efficient queries for containment (whether a key is in the map). Map-like
699 containers generally do not support efficient reverse mapping (values to
700 keys). If you need that, use two maps. Some map-like containers also
701 support efficient iteration through the keys in sorted order. Map-like
702 containers are the most expensive sort, only use them if you need one of
703 these capabilities.</li>
705 <li>a <a href="#ds_set">set-like</a> container if you need to put a bunch of
706 stuff into a container that automatically eliminates duplicates. Some
707 set-like containers support efficient iteration through the elements in
708 sorted order. Set-like containers are more expensive than sequential
712 <li>a <a href="#ds_sequential">sequential</a> container provides
713 the most efficient way to add elements and keeps track of the order they are
714 added to the collection. They permit duplicates and support efficient
715 iteration, but do not support efficient look-up based on a key.
718 <li>a <a href="#ds_bit">bit</a> container provides an efficient way to store and
719 perform set operations on sets of numeric id's, while automatically
720 eliminating duplicates. Bit containers require a maximum of 1 bit for each
721 identifier you want to store.
726 Once the proper category of container is determined, you can fine tune the
727 memory use, constant factors, and cache behaviors of access by intelligently
728 picking a member of the category. Note that constant factors and cache behavior
729 can be a big deal. If you have a vector that usually only contains a few
730 elements (but could contain many), for example, it's much better to use
731 <a href="#dss_smallvector">SmallVector</a> than <a href="#dss_vector">vector</a>
732 . Doing so avoids (relatively) expensive malloc/free calls, which dwarf the
733 cost of adding the elements to the container. </p>
737 <!-- ======================================================================= -->
738 <div class="doc_subsection">
739 <a name="ds_sequential">Sequential Containers (std::vector, std::list, etc)</a>
742 <div class="doc_text">
743 There are a variety of sequential containers available for you, based on your
744 needs. Pick the first in this section that will do what you want.
747 <!-- _______________________________________________________________________ -->
748 <div class="doc_subsubsection">
749 <a name="dss_fixedarrays">Fixed Size Arrays</a>
752 <div class="doc_text">
753 <p>Fixed size arrays are very simple and very fast. They are good if you know
754 exactly how many elements you have, or you have a (low) upper bound on how many
758 <!-- _______________________________________________________________________ -->
759 <div class="doc_subsubsection">
760 <a name="dss_heaparrays">Heap Allocated Arrays</a>
763 <div class="doc_text">
764 <p>Heap allocated arrays (new[] + delete[]) are also simple. They are good if
765 the number of elements is variable, if you know how many elements you will need
766 before the array is allocated, and if the array is usually large (if not,
767 consider a <a href="#dss_smallvector">SmallVector</a>). The cost of a heap
768 allocated array is the cost of the new/delete (aka malloc/free). Also note that
769 if you are allocating an array of a type with a constructor, the constructor and
770 destructors will be run for every element in the array (re-sizable vectors only
771 construct those elements actually used).</p>
774 <!-- _______________________________________________________________________ -->
775 <div class="doc_subsubsection">
776 <a name="dss_smallvector">"llvm/ADT/SmallVector.h"</a>
779 <div class="doc_text">
780 <p><tt>SmallVector<Type, N></tt> is a simple class that looks and smells
781 just like <tt>vector<Type></tt>:
782 it supports efficient iteration, lays out elements in memory order (so you can
783 do pointer arithmetic between elements), supports efficient push_back/pop_back
784 operations, supports efficient random access to its elements, etc.</p>
786 <p>The advantage of SmallVector is that it allocates space for
787 some number of elements (N) <b>in the object itself</b>. Because of this, if
788 the SmallVector is dynamically smaller than N, no malloc is performed. This can
789 be a big win in cases where the malloc/free call is far more expensive than the
790 code that fiddles around with the elements.</p>
792 <p>This is good for vectors that are "usually small" (e.g. the number of
793 predecessors/successors of a block is usually less than 8). On the other hand,
794 this makes the size of the SmallVector itself large, so you don't want to
795 allocate lots of them (doing so will waste a lot of space). As such,
796 SmallVectors are most useful when on the stack.</p>
798 <p>SmallVector also provides a nice portable and efficient replacement for
803 <!-- _______________________________________________________________________ -->
804 <div class="doc_subsubsection">
805 <a name="dss_vector"><vector></a>
808 <div class="doc_text">
810 std::vector is well loved and respected. It is useful when SmallVector isn't:
811 when the size of the vector is often large (thus the small optimization will
812 rarely be a benefit) or if you will be allocating many instances of the vector
813 itself (which would waste space for elements that aren't in the container).
814 vector is also useful when interfacing with code that expects vectors :).
817 <p>One worthwhile note about std::vector: avoid code like this:</p>
819 <div class="doc_code">
822 std::vector<foo> V;
828 <p>Instead, write this as:</p>
830 <div class="doc_code">
832 std::vector<foo> V;
840 <p>Doing so will save (at least) one heap allocation and free per iteration of
845 <!-- _______________________________________________________________________ -->
846 <div class="doc_subsubsection">
847 <a name="dss_deque"><deque></a>
850 <div class="doc_text">
851 <p>std::deque is, in some senses, a generalized version of std::vector. Like
852 std::vector, it provides constant time random access and other similar
853 properties, but it also provides efficient access to the front of the list. It
854 does not guarantee continuity of elements within memory.</p>
856 <p>In exchange for this extra flexibility, std::deque has significantly higher
857 constant factor costs than std::vector. If possible, use std::vector or
858 something cheaper.</p>
861 <!-- _______________________________________________________________________ -->
862 <div class="doc_subsubsection">
863 <a name="dss_list"><list></a>
866 <div class="doc_text">
867 <p>std::list is an extremely inefficient class that is rarely useful.
868 It performs a heap allocation for every element inserted into it, thus having an
869 extremely high constant factor, particularly for small data types. std::list
870 also only supports bidirectional iteration, not random access iteration.</p>
872 <p>In exchange for this high cost, std::list supports efficient access to both
873 ends of the list (like std::deque, but unlike std::vector or SmallVector). In
874 addition, the iterator invalidation characteristics of std::list are stronger
875 than that of a vector class: inserting or removing an element into the list does
876 not invalidate iterator or pointers to other elements in the list.</p>
879 <!-- _______________________________________________________________________ -->
880 <div class="doc_subsubsection">
881 <a name="dss_ilist">llvm/ADT/ilist</a>
884 <div class="doc_text">
885 <p><tt>ilist<T></tt> implements an 'intrusive' doubly-linked list. It is
886 intrusive, because it requires the element to store and provide access to the
887 prev/next pointers for the list.</p>
889 <p>ilist has the same drawbacks as std::list, and additionally requires an
890 ilist_traits implementation for the element type, but it provides some novel
891 characteristics. In particular, it can efficiently store polymorphic objects,
892 the traits class is informed when an element is inserted or removed from the
893 list, and ilists are guaranteed to support a constant-time splice operation.
896 <p>These properties are exactly what we want for things like Instructions and
897 basic blocks, which is why these are implemented with ilists.</p>
900 <!-- _______________________________________________________________________ -->
901 <div class="doc_subsubsection">
902 <a name="dss_other">Other Sequential Container options</a>
905 <div class="doc_text">
906 <p>Other STL containers are available, such as std::string.</p>
908 <p>There are also various STL adapter classes such as std::queue,
909 std::priority_queue, std::stack, etc. These provide simplified access to an
910 underlying container but don't affect the cost of the container itself.</p>
915 <!-- ======================================================================= -->
916 <div class="doc_subsection">
917 <a name="ds_set">Set-Like Containers (std::set, SmallSet, SetVector, etc)</a>
920 <div class="doc_text">
922 <p>Set-like containers are useful when you need to canonicalize multiple values
923 into a single representation. There are several different choices for how to do
924 this, providing various trade-offs.</p>
929 <!-- _______________________________________________________________________ -->
930 <div class="doc_subsubsection">
931 <a name="dss_sortedvectorset">A sorted 'vector'</a>
934 <div class="doc_text">
936 <p>If you intend to insert a lot of elements, then do a lot of queries, a
937 great approach is to use a vector (or other sequential container) with
938 std::sort+std::unique to remove duplicates. This approach works really well if
939 your usage pattern has these two distinct phases (insert then query), and can be
940 coupled with a good choice of <a href="#ds_sequential">sequential container</a>.
944 This combination provides the several nice properties: the result data is
945 contiguous in memory (good for cache locality), has few allocations, is easy to
946 address (iterators in the final vector are just indices or pointers), and can be
947 efficiently queried with a standard binary or radix search.</p>
951 <!-- _______________________________________________________________________ -->
952 <div class="doc_subsubsection">
953 <a name="dss_smallset">"llvm/ADT/SmallSet.h"</a>
956 <div class="doc_text">
958 <p>If you have a set-like data structure that is usually small and whose elements
959 are reasonably small, a <tt>SmallSet<Type, N></tt> is a good choice. This set
960 has space for N elements in place (thus, if the set is dynamically smaller than
961 N, no malloc traffic is required) and accesses them with a simple linear search.
962 When the set grows beyond 'N' elements, it allocates a more expensive representation that
963 guarantees efficient access (for most types, it falls back to std::set, but for
964 pointers it uses something far better, <a
965 href="#dss_smallptrset">SmallPtrSet</a>).</p>
967 <p>The magic of this class is that it handles small sets extremely efficiently,
968 but gracefully handles extremely large sets without loss of efficiency. The
969 drawback is that the interface is quite small: it supports insertion, queries
970 and erasing, but does not support iteration.</p>
974 <!-- _______________________________________________________________________ -->
975 <div class="doc_subsubsection">
976 <a name="dss_smallptrset">"llvm/ADT/SmallPtrSet.h"</a>
979 <div class="doc_text">
981 <p>SmallPtrSet has all the advantages of SmallSet (and a SmallSet of pointers is
982 transparently implemented with a SmallPtrSet), but also supports iterators. If
983 more than 'N' insertions are performed, a single quadratically
984 probed hash table is allocated and grows as needed, providing extremely
985 efficient access (constant time insertion/deleting/queries with low constant
986 factors) and is very stingy with malloc traffic.</p>
988 <p>Note that, unlike std::set, the iterators of SmallPtrSet are invalidated
989 whenever an insertion occurs. Also, the values visited by the iterators are not
990 visited in sorted order.</p>
994 <!-- _______________________________________________________________________ -->
995 <div class="doc_subsubsection">
996 <a name="dss_denseset">"llvm/ADT/DenseSet.h"</a>
999 <div class="doc_text">
1002 DenseSet is a simple quadratically probed hash table. It excels at supporting
1003 small values: it uses a single allocation to hold all of the pairs that
1004 are currently inserted in the set. DenseSet is a great way to unique small
1005 values that are not simple pointers (use <a
1006 href="#dss_smallptrset">SmallPtrSet</a> for pointers). Note that DenseSet has
1007 the same requirements for the value type that <a
1008 href="#dss_densemap">DenseMap</a> has.
1013 <!-- _______________________________________________________________________ -->
1014 <div class="doc_subsubsection">
1015 <a name="dss_FoldingSet">"llvm/ADT/FoldingSet.h"</a>
1018 <div class="doc_text">
1021 FoldingSet is an aggregate class that is really good at uniquing
1022 expensive-to-create or polymorphic objects. It is a combination of a chained
1023 hash table with intrusive links (uniqued objects are required to inherit from
1024 FoldingSetNode) that uses <a href="#dss_smallvector">SmallVector</a> as part of
1027 <p>Consider a case where you want to implement a "getOrCreateFoo" method for
1028 a complex object (for example, a node in the code generator). The client has a
1029 description of *what* it wants to generate (it knows the opcode and all the
1030 operands), but we don't want to 'new' a node, then try inserting it into a set
1031 only to find out it already exists, at which point we would have to delete it
1032 and return the node that already exists.
1035 <p>To support this style of client, FoldingSet perform a query with a
1036 FoldingSetNodeID (which wraps SmallVector) that can be used to describe the
1037 element that we want to query for. The query either returns the element
1038 matching the ID or it returns an opaque ID that indicates where insertion should
1039 take place. Construction of the ID usually does not require heap traffic.</p>
1041 <p>Because FoldingSet uses intrusive links, it can support polymorphic objects
1042 in the set (for example, you can have SDNode instances mixed with LoadSDNodes).
1043 Because the elements are individually allocated, pointers to the elements are
1044 stable: inserting or removing elements does not invalidate any pointers to other
1050 <!-- _______________________________________________________________________ -->
1051 <div class="doc_subsubsection">
1052 <a name="dss_set"><set></a>
1055 <div class="doc_text">
1057 <p><tt>std::set</tt> is a reasonable all-around set class, which is decent at
1058 many things but great at nothing. std::set allocates memory for each element
1059 inserted (thus it is very malloc intensive) and typically stores three pointers
1060 per element in the set (thus adding a large amount of per-element space
1061 overhead). It offers guaranteed log(n) performance, which is not particularly
1062 fast from a complexity standpoint (particularly if the elements of the set are
1063 expensive to compare, like strings), and has extremely high constant factors for
1064 lookup, insertion and removal.</p>
1066 <p>The advantages of std::set are that its iterators are stable (deleting or
1067 inserting an element from the set does not affect iterators or pointers to other
1068 elements) and that iteration over the set is guaranteed to be in sorted order.
1069 If the elements in the set are large, then the relative overhead of the pointers
1070 and malloc traffic is not a big deal, but if the elements of the set are small,
1071 std::set is almost never a good choice.</p>
1075 <!-- _______________________________________________________________________ -->
1076 <div class="doc_subsubsection">
1077 <a name="dss_setvector">"llvm/ADT/SetVector.h"</a>
1080 <div class="doc_text">
1081 <p>LLVM's SetVector<Type> is an adapter class that combines your choice of
1082 a set-like container along with a <a href="#ds_sequential">Sequential
1083 Container</a>. The important property
1084 that this provides is efficient insertion with uniquing (duplicate elements are
1085 ignored) with iteration support. It implements this by inserting elements into
1086 both a set-like container and the sequential container, using the set-like
1087 container for uniquing and the sequential container for iteration.
1090 <p>The difference between SetVector and other sets is that the order of
1091 iteration is guaranteed to match the order of insertion into the SetVector.
1092 This property is really important for things like sets of pointers. Because
1093 pointer values are non-deterministic (e.g. vary across runs of the program on
1094 different machines), iterating over the pointers in the set will
1095 not be in a well-defined order.</p>
1098 The drawback of SetVector is that it requires twice as much space as a normal
1099 set and has the sum of constant factors from the set-like container and the
1100 sequential container that it uses. Use it *only* if you need to iterate over
1101 the elements in a deterministic order. SetVector is also expensive to delete
1102 elements out of (linear time), unless you use it's "pop_back" method, which is
1106 <p>SetVector is an adapter class that defaults to using std::vector and std::set
1107 for the underlying containers, so it is quite expensive. However,
1108 <tt>"llvm/ADT/SetVector.h"</tt> also provides a SmallSetVector class, which
1109 defaults to using a SmallVector and SmallSet of a specified size. If you use
1110 this, and if your sets are dynamically smaller than N, you will save a lot of
1115 <!-- _______________________________________________________________________ -->
1116 <div class="doc_subsubsection">
1117 <a name="dss_uniquevector">"llvm/ADT/UniqueVector.h"</a>
1120 <div class="doc_text">
1123 UniqueVector is similar to <a href="#dss_setvector">SetVector</a>, but it
1124 retains a unique ID for each element inserted into the set. It internally
1125 contains a map and a vector, and it assigns a unique ID for each value inserted
1128 <p>UniqueVector is very expensive: its cost is the sum of the cost of
1129 maintaining both the map and vector, it has high complexity, high constant
1130 factors, and produces a lot of malloc traffic. It should be avoided.</p>
1135 <!-- _______________________________________________________________________ -->
1136 <div class="doc_subsubsection">
1137 <a name="dss_otherset">Other Set-Like Container Options</a>
1140 <div class="doc_text">
1143 The STL provides several other options, such as std::multiset and the various
1144 "hash_set" like containers (whether from C++ TR1 or from the SGI library).</p>
1146 <p>std::multiset is useful if you're not interested in elimination of
1147 duplicates, but has all the drawbacks of std::set. A sorted vector (where you
1148 don't delete duplicate entries) or some other approach is almost always
1151 <p>The various hash_set implementations (exposed portably by
1152 "llvm/ADT/hash_set") is a simple chained hashtable. This algorithm is as malloc
1153 intensive as std::set (performing an allocation for each element inserted,
1154 thus having really high constant factors) but (usually) provides O(1)
1155 insertion/deletion of elements. This can be useful if your elements are large
1156 (thus making the constant-factor cost relatively low) or if comparisons are
1157 expensive. Element iteration does not visit elements in a useful order.</p>
1161 <!-- ======================================================================= -->
1162 <div class="doc_subsection">
1163 <a name="ds_map">Map-Like Containers (std::map, DenseMap, etc)</a>
1166 <div class="doc_text">
1167 Map-like containers are useful when you want to associate data to a key. As
1168 usual, there are a lot of different ways to do this. :)
1171 <!-- _______________________________________________________________________ -->
1172 <div class="doc_subsubsection">
1173 <a name="dss_sortedvectormap">A sorted 'vector'</a>
1176 <div class="doc_text">
1179 If your usage pattern follows a strict insert-then-query approach, you can
1180 trivially use the same approach as <a href="#dss_sortedvectorset">sorted vectors
1181 for set-like containers</a>. The only difference is that your query function
1182 (which uses std::lower_bound to get efficient log(n) lookup) should only compare
1183 the key, not both the key and value. This yields the same advantages as sorted
1188 <!-- _______________________________________________________________________ -->
1189 <div class="doc_subsubsection">
1190 <a name="dss_stringmap">"llvm/ADT/StringMap.h"</a>
1193 <div class="doc_text">
1196 Strings are commonly used as keys in maps, and they are difficult to support
1197 efficiently: they are variable length, inefficient to hash and compare when
1198 long, expensive to copy, etc. StringMap is a specialized container designed to
1199 cope with these issues. It supports mapping an arbitrary range of bytes to an
1200 arbitrary other object.</p>
1202 <p>The StringMap implementation uses a quadratically-probed hash table, where
1203 the buckets store a pointer to the heap allocated entries (and some other
1204 stuff). The entries in the map must be heap allocated because the strings are
1205 variable length. The string data (key) and the element object (value) are
1206 stored in the same allocation with the string data immediately after the element
1207 object. This container guarantees the "<tt>(char*)(&Value+1)</tt>" points
1208 to the key string for a value.</p>
1210 <p>The StringMap is very fast for several reasons: quadratic probing is very
1211 cache efficient for lookups, the hash value of strings in buckets is not
1212 recomputed when lookup up an element, StringMap rarely has to touch the
1213 memory for unrelated objects when looking up a value (even when hash collisions
1214 happen), hash table growth does not recompute the hash values for strings
1215 already in the table, and each pair in the map is store in a single allocation
1216 (the string data is stored in the same allocation as the Value of a pair).</p>
1218 <p>StringMap also provides query methods that take byte ranges, so it only ever
1219 copies a string if a value is inserted into the table.</p>
1222 <!-- _______________________________________________________________________ -->
1223 <div class="doc_subsubsection">
1224 <a name="dss_indexedmap">"llvm/ADT/IndexedMap.h"</a>
1227 <div class="doc_text">
1229 IndexedMap is a specialized container for mapping small dense integers (or
1230 values that can be mapped to small dense integers) to some other type. It is
1231 internally implemented as a vector with a mapping function that maps the keys to
1232 the dense integer range.
1236 This is useful for cases like virtual registers in the LLVM code generator: they
1237 have a dense mapping that is offset by a compile-time constant (the first
1238 virtual register ID).</p>
1242 <!-- _______________________________________________________________________ -->
1243 <div class="doc_subsubsection">
1244 <a name="dss_densemap">"llvm/ADT/DenseMap.h"</a>
1247 <div class="doc_text">
1250 DenseMap is a simple quadratically probed hash table. It excels at supporting
1251 small keys and values: it uses a single allocation to hold all of the pairs that
1252 are currently inserted in the map. DenseMap is a great way to map pointers to
1253 pointers, or map other small types to each other.
1257 There are several aspects of DenseMap that you should be aware of, however. The
1258 iterators in a densemap are invalidated whenever an insertion occurs, unlike
1259 map. Also, because DenseMap allocates space for a large number of key/value
1260 pairs (it starts with 64 by default), it will waste a lot of space if your keys
1261 or values are large. Finally, you must implement a partial specialization of
1262 DenseMapInfo for the key that you want, if it isn't already supported. This
1263 is required to tell DenseMap about two special marker values (which can never be
1264 inserted into the map) that it needs internally.</p>
1268 <!-- _______________________________________________________________________ -->
1269 <div class="doc_subsubsection">
1270 <a name="dss_map"><map></a>
1273 <div class="doc_text">
1276 std::map has similar characteristics to <a href="#dss_set">std::set</a>: it uses
1277 a single allocation per pair inserted into the map, it offers log(n) lookup with
1278 an extremely large constant factor, imposes a space penalty of 3 pointers per
1279 pair in the map, etc.</p>
1281 <p>std::map is most useful when your keys or values are very large, if you need
1282 to iterate over the collection in sorted order, or if you need stable iterators
1283 into the map (i.e. they don't get invalidated if an insertion or deletion of
1284 another element takes place).</p>
1288 <!-- _______________________________________________________________________ -->
1289 <div class="doc_subsubsection">
1290 <a name="dss_othermap">Other Map-Like Container Options</a>
1293 <div class="doc_text">
1296 The STL provides several other options, such as std::multimap and the various
1297 "hash_map" like containers (whether from C++ TR1 or from the SGI library).</p>
1299 <p>std::multimap is useful if you want to map a key to multiple values, but has
1300 all the drawbacks of std::map. A sorted vector or some other approach is almost
1303 <p>The various hash_map implementations (exposed portably by
1304 "llvm/ADT/hash_map") are simple chained hash tables. This algorithm is as
1305 malloc intensive as std::map (performing an allocation for each element
1306 inserted, thus having really high constant factors) but (usually) provides O(1)
1307 insertion/deletion of elements. This can be useful if your elements are large
1308 (thus making the constant-factor cost relatively low) or if comparisons are
1309 expensive. Element iteration does not visit elements in a useful order.</p>
1313 <!-- ======================================================================= -->
1314 <div class="doc_subsection">
1315 <a name="ds_bit">Bit storage containers (BitVector, SparseBitVector)</a>
1318 <div class="doc_text">
1319 <p>Unlike the other containers, there are only two bit storage containers, and
1320 choosing when to use each is relatively straightforward.</p>
1322 <p>One additional option is
1323 <tt>std::vector<bool></tt>: we discourage its use for two reasons 1) the
1324 implementation in many common compilers (e.g. commonly available versions of
1325 GCC) is extremely inefficient and 2) the C++ standards committee is likely to
1326 deprecate this container and/or change it significantly somehow. In any case,
1327 please don't use it.</p>
1330 <!-- _______________________________________________________________________ -->
1331 <div class="doc_subsubsection">
1332 <a name="dss_bitvector">BitVector</a>
1335 <div class="doc_text">
1336 <p> The BitVector container provides a fixed size set of bits for manipulation.
1337 It supports individual bit setting/testing, as well as set operations. The set
1338 operations take time O(size of bitvector), but operations are performed one word
1339 at a time, instead of one bit at a time. This makes the BitVector very fast for
1340 set operations compared to other containers. Use the BitVector when you expect
1341 the number of set bits to be high (IE a dense set).
1345 <!-- _______________________________________________________________________ -->
1346 <div class="doc_subsubsection">
1347 <a name="dss_sparsebitvector">SparseBitVector</a>
1350 <div class="doc_text">
1351 <p> The SparseBitVector container is much like BitVector, with one major
1352 difference: Only the bits that are set, are stored. This makes the
1353 SparseBitVector much more space efficient than BitVector when the set is sparse,
1354 as well as making set operations O(number of set bits) instead of O(size of
1355 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
1356 (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).
1360 <!-- *********************************************************************** -->
1361 <div class="doc_section">
1362 <a name="common">Helpful Hints for Common Operations</a>
1364 <!-- *********************************************************************** -->
1366 <div class="doc_text">
1368 <p>This section describes how to perform some very simple transformations of
1369 LLVM code. This is meant to give examples of common idioms used, showing the
1370 practical side of LLVM transformations. <p> Because this is a "how-to" section,
1371 you should also read about the main classes that you will be working with. The
1372 <a href="#coreclasses">Core LLVM Class Hierarchy Reference</a> contains details
1373 and descriptions of the main classes that you should know about.</p>
1377 <!-- NOTE: this section should be heavy on example code -->
1378 <!-- ======================================================================= -->
1379 <div class="doc_subsection">
1380 <a name="inspection">Basic Inspection and Traversal Routines</a>
1383 <div class="doc_text">
1385 <p>The LLVM compiler infrastructure have many different data structures that may
1386 be traversed. Following the example of the C++ standard template library, the
1387 techniques used to traverse these various data structures are all basically the
1388 same. For a enumerable sequence of values, the <tt>XXXbegin()</tt> function (or
1389 method) returns an iterator to the start of the sequence, the <tt>XXXend()</tt>
1390 function returns an iterator pointing to one past the last valid element of the
1391 sequence, and there is some <tt>XXXiterator</tt> data type that is common
1392 between the two operations.</p>
1394 <p>Because the pattern for iteration is common across many different aspects of
1395 the program representation, the standard template library algorithms may be used
1396 on them, and it is easier to remember how to iterate. First we show a few common
1397 examples of the data structures that need to be traversed. Other data
1398 structures are traversed in very similar ways.</p>
1402 <!-- _______________________________________________________________________ -->
1403 <div class="doc_subsubsection">
1404 <a name="iterate_function">Iterating over the </a><a
1405 href="#BasicBlock"><tt>BasicBlock</tt></a>s in a <a
1406 href="#Function"><tt>Function</tt></a>
1409 <div class="doc_text">
1411 <p>It's quite common to have a <tt>Function</tt> instance that you'd like to
1412 transform in some way; in particular, you'd like to manipulate its
1413 <tt>BasicBlock</tt>s. To facilitate this, you'll need to iterate over all of
1414 the <tt>BasicBlock</tt>s that constitute the <tt>Function</tt>. The following is
1415 an example that prints the name of a <tt>BasicBlock</tt> and the number of
1416 <tt>Instruction</tt>s it contains:</p>
1418 <div class="doc_code">
1420 // <i>func is a pointer to a Function instance</i>
1421 for (Function::iterator i = func->begin(), e = func->end(); i != e; ++i)
1422 // <i>Print out the name of the basic block if it has one, and then the</i>
1423 // <i>number of instructions that it contains</i>
1424 llvm::cerr << "Basic block (name=" << i->getName() << ") has "
1425 << i->size() << " instructions.\n";
1429 <p>Note that i can be used as if it were a pointer for the purposes of
1430 invoking member functions of the <tt>Instruction</tt> class. This is
1431 because the indirection operator is overloaded for the iterator
1432 classes. In the above code, the expression <tt>i->size()</tt> is
1433 exactly equivalent to <tt>(*i).size()</tt> just like you'd expect.</p>
1437 <!-- _______________________________________________________________________ -->
1438 <div class="doc_subsubsection">
1439 <a name="iterate_basicblock">Iterating over the </a><a
1440 href="#Instruction"><tt>Instruction</tt></a>s in a <a
1441 href="#BasicBlock"><tt>BasicBlock</tt></a>
1444 <div class="doc_text">
1446 <p>Just like when dealing with <tt>BasicBlock</tt>s in <tt>Function</tt>s, it's
1447 easy to iterate over the individual instructions that make up
1448 <tt>BasicBlock</tt>s. Here's a code snippet that prints out each instruction in
1449 a <tt>BasicBlock</tt>:</p>
1451 <div class="doc_code">
1453 // <i>blk is a pointer to a BasicBlock instance</i>
1454 for (BasicBlock::iterator i = blk->begin(), e = blk->end(); i != e; ++i)
1455 // <i>The next statement works since operator<<(ostream&,...)</i>
1456 // <i>is overloaded for Instruction&</i>
1457 llvm::cerr << *i << "\n";
1461 <p>However, this isn't really the best way to print out the contents of a
1462 <tt>BasicBlock</tt>! Since the ostream operators are overloaded for virtually
1463 anything you'll care about, you could have just invoked the print routine on the
1464 basic block itself: <tt>llvm::cerr << *blk << "\n";</tt>.</p>
1468 <!-- _______________________________________________________________________ -->
1469 <div class="doc_subsubsection">
1470 <a name="iterate_institer">Iterating over the </a><a
1471 href="#Instruction"><tt>Instruction</tt></a>s in a <a
1472 href="#Function"><tt>Function</tt></a>
1475 <div class="doc_text">
1477 <p>If you're finding that you commonly iterate over a <tt>Function</tt>'s
1478 <tt>BasicBlock</tt>s and then that <tt>BasicBlock</tt>'s <tt>Instruction</tt>s,
1479 <tt>InstIterator</tt> should be used instead. You'll need to include <a
1480 href="/doxygen/InstIterator_8h-source.html"><tt>llvm/Support/InstIterator.h</tt></a>,
1481 and then instantiate <tt>InstIterator</tt>s explicitly in your code. Here's a
1482 small example that shows how to dump all instructions in a function to the standard error stream:<p>
1484 <div class="doc_code">
1486 #include "<a href="/doxygen/InstIterator_8h-source.html">llvm/Support/InstIterator.h</a>"
1488 // <i>F is a pointer to a Function instance</i>
1489 for (inst_iterator I = inst_begin(F), E = inst_end(F); I != E; ++I)
1490 llvm::cerr << *I << "\n";
1494 <p>Easy, isn't it? You can also use <tt>InstIterator</tt>s to fill a
1495 work list with its initial contents. For example, if you wanted to
1496 initialize a work list to contain all instructions in a <tt>Function</tt>
1497 F, all you would need to do is something like:</p>
1499 <div class="doc_code">
1501 std::set<Instruction*> worklist;
1502 // or better yet, SmallPtrSet<Instruction*, 64> worklist;
1504 for (inst_iterator I = inst_begin(F), E = inst_end(F); I != E; ++I)
1505 worklist.insert(&*I);
1509 <p>The STL set <tt>worklist</tt> would now contain all instructions in the
1510 <tt>Function</tt> pointed to by F.</p>
1514 <!-- _______________________________________________________________________ -->
1515 <div class="doc_subsubsection">
1516 <a name="iterate_convert">Turning an iterator into a class pointer (and
1520 <div class="doc_text">
1522 <p>Sometimes, it'll be useful to grab a reference (or pointer) to a class
1523 instance when all you've got at hand is an iterator. Well, extracting
1524 a reference or a pointer from an iterator is very straight-forward.
1525 Assuming that <tt>i</tt> is a <tt>BasicBlock::iterator</tt> and <tt>j</tt>
1526 is a <tt>BasicBlock::const_iterator</tt>:</p>
1528 <div class="doc_code">
1530 Instruction& inst = *i; // <i>Grab reference to instruction reference</i>
1531 Instruction* pinst = &*i; // <i>Grab pointer to instruction reference</i>
1532 const Instruction& inst = *j;
1536 <p>However, the iterators you'll be working with in the LLVM framework are
1537 special: they will automatically convert to a ptr-to-instance type whenever they
1538 need to. Instead of dereferencing the iterator and then taking the address of
1539 the result, you can simply assign the iterator to the proper pointer type and
1540 you get the dereference and address-of operation as a result of the assignment
1541 (behind the scenes, this is a result of overloading casting mechanisms). Thus
1542 the last line of the last example,</p>
1544 <div class="doc_code">
1546 Instruction *pinst = &*i;
1550 <p>is semantically equivalent to</p>
1552 <div class="doc_code">
1554 Instruction *pinst = i;
1558 <p>It's also possible to turn a class pointer into the corresponding iterator,
1559 and this is a constant time operation (very efficient). The following code
1560 snippet illustrates use of the conversion constructors provided by LLVM
1561 iterators. By using these, you can explicitly grab the iterator of something
1562 without actually obtaining it via iteration over some structure:</p>
1564 <div class="doc_code">
1566 void printNextInstruction(Instruction* inst) {
1567 BasicBlock::iterator it(inst);
1568 ++it; // <i>After this line, it refers to the instruction after *inst</i>
1569 if (it != inst->getParent()->end()) llvm::cerr << *it << "\n";
1576 <!--_______________________________________________________________________-->
1577 <div class="doc_subsubsection">
1578 <a name="iterate_complex">Finding call sites: a slightly more complex
1582 <div class="doc_text">
1584 <p>Say that you're writing a FunctionPass and would like to count all the
1585 locations in the entire module (that is, across every <tt>Function</tt>) where a
1586 certain function (i.e., some <tt>Function</tt>*) is already in scope. As you'll
1587 learn later, you may want to use an <tt>InstVisitor</tt> to accomplish this in a
1588 much more straight-forward manner, but this example will allow us to explore how
1589 you'd do it if you didn't have <tt>InstVisitor</tt> around. In pseudo-code, this
1590 is what we want to do:</p>
1592 <div class="doc_code">
1594 initialize callCounter to zero
1595 for each Function f in the Module
1596 for each BasicBlock b in f
1597 for each Instruction i in b
1598 if (i is a CallInst and calls the given function)
1599 increment callCounter
1603 <p>And the actual code is (remember, because we're writing a
1604 <tt>FunctionPass</tt>, our <tt>FunctionPass</tt>-derived class simply has to
1605 override the <tt>runOnFunction</tt> method):</p>
1607 <div class="doc_code">
1609 Function* targetFunc = ...;
1611 class OurFunctionPass : public FunctionPass {
1613 OurFunctionPass(): callCounter(0) { }
1615 virtual runOnFunction(Function& F) {
1616 for (Function::iterator b = F.begin(), be = F.end(); b != be; ++b) {
1617 for (BasicBlock::iterator i = b->begin(), ie = b->end(); i != ie; ++i) {
1618 if (<a href="#CallInst">CallInst</a>* callInst = <a href="#isa">dyn_cast</a><<a
1619 href="#CallInst">CallInst</a>>(&*i)) {
1620 // <i>We know we've encountered a call instruction, so we</i>
1621 // <i>need to determine if it's a call to the</i>
1622 // <i>function pointed to by m_func or not.</i>
1623 if (callInst->getCalledFunction() == targetFunc)
1631 unsigned callCounter;
1638 <!--_______________________________________________________________________-->
1639 <div class="doc_subsubsection">
1640 <a name="calls_and_invokes">Treating calls and invokes the same way</a>
1643 <div class="doc_text">
1645 <p>You may have noticed that the previous example was a bit oversimplified in
1646 that it did not deal with call sites generated by 'invoke' instructions. In
1647 this, and in other situations, you may find that you want to treat
1648 <tt>CallInst</tt>s and <tt>InvokeInst</tt>s the same way, even though their
1649 most-specific common base class is <tt>Instruction</tt>, which includes lots of
1650 less closely-related things. For these cases, LLVM provides a handy wrapper
1652 href="http://llvm.org/doxygen/classllvm_1_1CallSite.html"><tt>CallSite</tt></a>.
1653 It is essentially a wrapper around an <tt>Instruction</tt> pointer, with some
1654 methods that provide functionality common to <tt>CallInst</tt>s and
1655 <tt>InvokeInst</tt>s.</p>
1657 <p>This class has "value semantics": it should be passed by value, not by
1658 reference and it should not be dynamically allocated or deallocated using
1659 <tt>operator new</tt> or <tt>operator delete</tt>. It is efficiently copyable,
1660 assignable and constructable, with costs equivalents to that of a bare pointer.
1661 If you look at its definition, it has only a single pointer member.</p>
1665 <!--_______________________________________________________________________-->
1666 <div class="doc_subsubsection">
1667 <a name="iterate_chains">Iterating over def-use & use-def chains</a>
1670 <div class="doc_text">
1672 <p>Frequently, we might have an instance of the <a
1673 href="/doxygen/classllvm_1_1Value.html">Value Class</a> and we want to
1674 determine which <tt>User</tt>s use the <tt>Value</tt>. The list of all
1675 <tt>User</tt>s of a particular <tt>Value</tt> is called a <i>def-use</i> chain.
1676 For example, let's say we have a <tt>Function*</tt> named <tt>F</tt> to a
1677 particular function <tt>foo</tt>. Finding all of the instructions that
1678 <i>use</i> <tt>foo</tt> is as simple as iterating over the <i>def-use</i> chain
1681 <div class="doc_code">
1685 for (Value::use_iterator i = F->use_begin(), e = F->use_end(); i != e; ++i)
1686 if (Instruction *Inst = dyn_cast<Instruction>(*i)) {
1687 llvm::cerr << "F is used in instruction:\n";
1688 llvm::cerr << *Inst << "\n";
1693 <p>Alternately, it's common to have an instance of the <a
1694 href="/doxygen/classllvm_1_1User.html">User Class</a> and need to know what
1695 <tt>Value</tt>s are used by it. The list of all <tt>Value</tt>s used by a
1696 <tt>User</tt> is known as a <i>use-def</i> chain. Instances of class
1697 <tt>Instruction</tt> are common <tt>User</tt>s, so we might want to iterate over
1698 all of the values that a particular instruction uses (that is, the operands of
1699 the particular <tt>Instruction</tt>):</p>
1701 <div class="doc_code">
1703 Instruction *pi = ...;
1705 for (User::op_iterator i = pi->op_begin(), e = pi->op_end(); i != e; ++i) {
1713 def-use chains ("finding all users of"): Value::use_begin/use_end
1714 use-def chains ("finding all values used"): User::op_begin/op_end [op=operand]
1719 <!--_______________________________________________________________________-->
1720 <div class="doc_subsubsection">
1721 <a name="iterate_preds">Iterating over predecessors &
1722 successors of blocks</a>
1725 <div class="doc_text">
1727 <p>Iterating over the predecessors and successors of a block is quite easy
1728 with the routines defined in <tt>"llvm/Support/CFG.h"</tt>. Just use code like
1729 this to iterate over all predecessors of BB:</p>
1731 <div class="doc_code">
1733 #include "llvm/Support/CFG.h"
1734 BasicBlock *BB = ...;
1736 for (pred_iterator PI = pred_begin(BB), E = pred_end(BB); PI != E; ++PI) {
1737 BasicBlock *Pred = *PI;
1743 <p>Similarly, to iterate over successors use
1744 succ_iterator/succ_begin/succ_end.</p>
1749 <!-- ======================================================================= -->
1750 <div class="doc_subsection">
1751 <a name="simplechanges">Making simple changes</a>
1754 <div class="doc_text">
1756 <p>There are some primitive transformation operations present in the LLVM
1757 infrastructure that are worth knowing about. When performing
1758 transformations, it's fairly common to manipulate the contents of basic
1759 blocks. This section describes some of the common methods for doing so
1760 and gives example code.</p>
1764 <!--_______________________________________________________________________-->
1765 <div class="doc_subsubsection">
1766 <a name="schanges_creating">Creating and inserting new
1767 <tt>Instruction</tt>s</a>
1770 <div class="doc_text">
1772 <p><i>Instantiating Instructions</i></p>
1774 <p>Creation of <tt>Instruction</tt>s is straight-forward: simply call the
1775 constructor for the kind of instruction to instantiate and provide the necessary
1776 parameters. For example, an <tt>AllocaInst</tt> only <i>requires</i> a
1777 (const-ptr-to) <tt>Type</tt>. Thus:</p>
1779 <div class="doc_code">
1781 AllocaInst* ai = new AllocaInst(Type::Int32Ty);
1785 <p>will create an <tt>AllocaInst</tt> instance that represents the allocation of
1786 one integer in the current stack frame, at run time. Each <tt>Instruction</tt>
1787 subclass is likely to have varying default parameters which change the semantics
1788 of the instruction, so refer to the <a
1789 href="/doxygen/classllvm_1_1Instruction.html">doxygen documentation for the subclass of
1790 Instruction</a> that you're interested in instantiating.</p>
1792 <p><i>Naming values</i></p>
1794 <p>It is very useful to name the values of instructions when you're able to, as
1795 this facilitates the debugging of your transformations. If you end up looking
1796 at generated LLVM machine code, you definitely want to have logical names
1797 associated with the results of instructions! By supplying a value for the
1798 <tt>Name</tt> (default) parameter of the <tt>Instruction</tt> constructor, you
1799 associate a logical name with the result of the instruction's execution at
1800 run time. For example, say that I'm writing a transformation that dynamically
1801 allocates space for an integer on the stack, and that integer is going to be
1802 used as some kind of index by some other code. To accomplish this, I place an
1803 <tt>AllocaInst</tt> at the first point in the first <tt>BasicBlock</tt> of some
1804 <tt>Function</tt>, and I'm intending to use it within the same
1805 <tt>Function</tt>. I might do:</p>
1807 <div class="doc_code">
1809 AllocaInst* pa = new AllocaInst(Type::Int32Ty, 0, "indexLoc");
1813 <p>where <tt>indexLoc</tt> is now the logical name of the instruction's
1814 execution value, which is a pointer to an integer on the run time stack.</p>
1816 <p><i>Inserting instructions</i></p>
1818 <p>There are essentially two ways to insert an <tt>Instruction</tt>
1819 into an existing sequence of instructions that form a <tt>BasicBlock</tt>:</p>
1822 <li>Insertion into an explicit instruction list
1824 <p>Given a <tt>BasicBlock* pb</tt>, an <tt>Instruction* pi</tt> within that
1825 <tt>BasicBlock</tt>, and a newly-created instruction we wish to insert
1826 before <tt>*pi</tt>, we do the following: </p>
1828 <div class="doc_code">
1830 BasicBlock *pb = ...;
1831 Instruction *pi = ...;
1832 Instruction *newInst = new Instruction(...);
1834 pb->getInstList().insert(pi, newInst); // <i>Inserts newInst before pi in pb</i>
1838 <p>Appending to the end of a <tt>BasicBlock</tt> is so common that
1839 the <tt>Instruction</tt> class and <tt>Instruction</tt>-derived
1840 classes provide constructors which take a pointer to a
1841 <tt>BasicBlock</tt> to be appended to. For example code that
1844 <div class="doc_code">
1846 BasicBlock *pb = ...;
1847 Instruction *newInst = new Instruction(...);
1849 pb->getInstList().push_back(newInst); // <i>Appends newInst to pb</i>
1855 <div class="doc_code">
1857 BasicBlock *pb = ...;
1858 Instruction *newInst = new Instruction(..., pb);
1862 <p>which is much cleaner, especially if you are creating
1863 long instruction streams.</p></li>
1865 <li>Insertion into an implicit instruction list
1867 <p><tt>Instruction</tt> instances that are already in <tt>BasicBlock</tt>s
1868 are implicitly associated with an existing instruction list: the instruction
1869 list of the enclosing basic block. Thus, we could have accomplished the same
1870 thing as the above code without being given a <tt>BasicBlock</tt> by doing:
1873 <div class="doc_code">
1875 Instruction *pi = ...;
1876 Instruction *newInst = new Instruction(...);
1878 pi->getParent()->getInstList().insert(pi, newInst);
1882 <p>In fact, this sequence of steps occurs so frequently that the
1883 <tt>Instruction</tt> class and <tt>Instruction</tt>-derived classes provide
1884 constructors which take (as a default parameter) a pointer to an
1885 <tt>Instruction</tt> which the newly-created <tt>Instruction</tt> should
1886 precede. That is, <tt>Instruction</tt> constructors are capable of
1887 inserting the newly-created instance into the <tt>BasicBlock</tt> of a
1888 provided instruction, immediately before that instruction. Using an
1889 <tt>Instruction</tt> constructor with a <tt>insertBefore</tt> (default)
1890 parameter, the above code becomes:</p>
1892 <div class="doc_code">
1894 Instruction* pi = ...;
1895 Instruction* newInst = new Instruction(..., pi);
1899 <p>which is much cleaner, especially if you're creating a lot of
1900 instructions and adding them to <tt>BasicBlock</tt>s.</p></li>
1905 <!--_______________________________________________________________________-->
1906 <div class="doc_subsubsection">
1907 <a name="schanges_deleting">Deleting <tt>Instruction</tt>s</a>
1910 <div class="doc_text">
1912 <p>Deleting an instruction from an existing sequence of instructions that form a
1913 <a href="#BasicBlock"><tt>BasicBlock</tt></a> is very straight-forward. First,
1914 you must have a pointer to the instruction that you wish to delete. Second, you
1915 need to obtain the pointer to that instruction's basic block. You use the
1916 pointer to the basic block to get its list of instructions and then use the
1917 erase function to remove your instruction. For example:</p>
1919 <div class="doc_code">
1921 <a href="#Instruction">Instruction</a> *I = .. ;
1922 I->eraseFromParent();
1928 <!--_______________________________________________________________________-->
1929 <div class="doc_subsubsection">
1930 <a name="schanges_replacing">Replacing an <tt>Instruction</tt> with another
1934 <div class="doc_text">
1936 <p><i>Replacing individual instructions</i></p>
1938 <p>Including "<a href="/doxygen/BasicBlockUtils_8h-source.html">llvm/Transforms/Utils/BasicBlockUtils.h</a>"
1939 permits use of two very useful replace functions: <tt>ReplaceInstWithValue</tt>
1940 and <tt>ReplaceInstWithInst</tt>.</p>
1942 <h4><a name="schanges_deleting">Deleting <tt>Instruction</tt>s</a></h4>
1945 <li><tt>ReplaceInstWithValue</tt>
1947 <p>This function replaces all uses of a given instruction with a value,
1948 and then removes the original instruction. The following example
1949 illustrates the replacement of the result of a particular
1950 <tt>AllocaInst</tt> that allocates memory for a single integer with a null
1951 pointer to an integer.</p>
1953 <div class="doc_code">
1955 AllocaInst* instToReplace = ...;
1956 BasicBlock::iterator ii(instToReplace);
1958 ReplaceInstWithValue(instToReplace->getParent()->getInstList(), ii,
1959 Constant::getNullValue(PointerType::getUnqual(Type::Int32Ty)));
1962 <li><tt>ReplaceInstWithInst</tt>
1964 <p>This function replaces a particular instruction with another
1965 instruction, inserting the new instruction into the basic block at the
1966 location where the old instruction was, and replacing any uses of the old
1967 instruction with the new instruction. The following example illustrates
1968 the replacement of one <tt>AllocaInst</tt> with another.</p>
1970 <div class="doc_code">
1972 AllocaInst* instToReplace = ...;
1973 BasicBlock::iterator ii(instToReplace);
1975 ReplaceInstWithInst(instToReplace->getParent()->getInstList(), ii,
1976 new AllocaInst(Type::Int32Ty, 0, "ptrToReplacedInt"));
1980 <p><i>Replacing multiple uses of <tt>User</tt>s and <tt>Value</tt>s</i></p>
1982 <p>You can use <tt>Value::replaceAllUsesWith</tt> and
1983 <tt>User::replaceUsesOfWith</tt> to change more than one use at a time. See the
1984 doxygen documentation for the <a href="/doxygen/classllvm_1_1Value.html">Value Class</a>
1985 and <a href="/doxygen/classllvm_1_1User.html">User Class</a>, respectively, for more
1988 <!-- Value::replaceAllUsesWith User::replaceUsesOfWith Point out:
1989 include/llvm/Transforms/Utils/ especially BasicBlockUtils.h with:
1990 ReplaceInstWithValue, ReplaceInstWithInst -->
1994 <!--_______________________________________________________________________-->
1995 <div class="doc_subsubsection">
1996 <a name="schanges_deletingGV">Deleting <tt>GlobalVariable</tt>s</a>
1999 <div class="doc_text">
2001 <p>Deleting a global variable from a module is just as easy as deleting an
2002 Instruction. First, you must have a pointer to the global variable that you wish
2003 to delete. You use this pointer to erase it from its parent, the module.
2006 <div class="doc_code">
2008 <a href="#GlobalVariable">GlobalVariable</a> *GV = .. ;
2010 GV->eraseFromParent();
2016 <!-- *********************************************************************** -->
2017 <div class="doc_section">
2018 <a name="advanced">Advanced Topics</a>
2020 <!-- *********************************************************************** -->
2022 <div class="doc_text">
2024 This section describes some of the advanced or obscure API's that most clients
2025 do not need to be aware of. These API's tend manage the inner workings of the
2026 LLVM system, and only need to be accessed in unusual circumstances.
2030 <!-- ======================================================================= -->
2031 <div class="doc_subsection">
2032 <a name="TypeResolve">LLVM Type Resolution</a>
2035 <div class="doc_text">
2038 The LLVM type system has a very simple goal: allow clients to compare types for
2039 structural equality with a simple pointer comparison (aka a shallow compare).
2040 This goal makes clients much simpler and faster, and is used throughout the LLVM
2045 Unfortunately achieving this goal is not a simple matter. In particular,
2046 recursive types and late resolution of opaque types makes the situation very
2047 difficult to handle. Fortunately, for the most part, our implementation makes
2048 most clients able to be completely unaware of the nasty internal details. The
2049 primary case where clients are exposed to the inner workings of it are when
2050 building a recursive type. In addition to this case, the LLVM bitcode reader,
2051 assembly parser, and linker also have to be aware of the inner workings of this
2056 For our purposes below, we need three concepts. First, an "Opaque Type" is
2057 exactly as defined in the <a href="LangRef.html#t_opaque">language
2058 reference</a>. Second an "Abstract Type" is any type which includes an
2059 opaque type as part of its type graph (for example "<tt>{ opaque, i32 }</tt>").
2060 Third, a concrete type is a type that is not an abstract type (e.g. "<tt>{ i32,
2066 <!-- ______________________________________________________________________ -->
2067 <div class="doc_subsubsection">
2068 <a name="BuildRecType">Basic Recursive Type Construction</a>
2071 <div class="doc_text">
2074 Because the most common question is "how do I build a recursive type with LLVM",
2075 we answer it now and explain it as we go. Here we include enough to cause this
2076 to be emitted to an output .ll file:
2079 <div class="doc_code">
2081 %mylist = type { %mylist*, i32 }
2086 To build this, use the following LLVM APIs:
2089 <div class="doc_code">
2091 // <i>Create the initial outer struct</i>
2092 <a href="#PATypeHolder">PATypeHolder</a> StructTy = OpaqueType::get();
2093 std::vector<const Type*> Elts;
2094 Elts.push_back(PointerType::getUnqual(StructTy));
2095 Elts.push_back(Type::Int32Ty);
2096 StructType *NewSTy = StructType::get(Elts);
2098 // <i>At this point, NewSTy = "{ opaque*, i32 }". Tell VMCore that</i>
2099 // <i>the struct and the opaque type are actually the same.</i>
2100 cast<OpaqueType>(StructTy.get())-><a href="#refineAbstractTypeTo">refineAbstractTypeTo</a>(NewSTy);
2102 // <i>NewSTy is potentially invalidated, but StructTy (a <a href="#PATypeHolder">PATypeHolder</a>) is</i>
2103 // <i>kept up-to-date</i>
2104 NewSTy = cast<StructType>(StructTy.get());
2106 // <i>Add a name for the type to the module symbol table (optional)</i>
2107 MyModule->addTypeName("mylist", NewSTy);
2112 This code shows the basic approach used to build recursive types: build a
2113 non-recursive type using 'opaque', then use type unification to close the cycle.
2114 The type unification step is performed by the <tt><a
2115 href="#refineAbstractTypeTo">refineAbstractTypeTo</a></tt> method, which is
2116 described next. After that, we describe the <a
2117 href="#PATypeHolder">PATypeHolder class</a>.
2122 <!-- ______________________________________________________________________ -->
2123 <div class="doc_subsubsection">
2124 <a name="refineAbstractTypeTo">The <tt>refineAbstractTypeTo</tt> method</a>
2127 <div class="doc_text">
2129 The <tt>refineAbstractTypeTo</tt> method starts the type unification process.
2130 While this method is actually a member of the DerivedType class, it is most
2131 often used on OpaqueType instances. Type unification is actually a recursive
2132 process. After unification, types can become structurally isomorphic to
2133 existing types, and all duplicates are deleted (to preserve pointer equality).
2137 In the example above, the OpaqueType object is definitely deleted.
2138 Additionally, if there is an "{ \2*, i32}" type already created in the system,
2139 the pointer and struct type created are <b>also</b> deleted. Obviously whenever
2140 a type is deleted, any "Type*" pointers in the program are invalidated. As
2141 such, it is safest to avoid having <i>any</i> "Type*" pointers to abstract types
2142 live across a call to <tt>refineAbstractTypeTo</tt> (note that non-abstract
2143 types can never move or be deleted). To deal with this, the <a
2144 href="#PATypeHolder">PATypeHolder</a> class is used to maintain a stable
2145 reference to a possibly refined type, and the <a
2146 href="#AbstractTypeUser">AbstractTypeUser</a> class is used to update more
2147 complex datastructures.
2152 <!-- ______________________________________________________________________ -->
2153 <div class="doc_subsubsection">
2154 <a name="PATypeHolder">The PATypeHolder Class</a>
2157 <div class="doc_text">
2159 PATypeHolder is a form of a "smart pointer" for Type objects. When VMCore
2160 happily goes about nuking types that become isomorphic to existing types, it
2161 automatically updates all PATypeHolder objects to point to the new type. In the
2162 example above, this allows the code to maintain a pointer to the resultant
2163 resolved recursive type, even though the Type*'s are potentially invalidated.
2167 PATypeHolder is an extremely light-weight object that uses a lazy union-find
2168 implementation to update pointers. For example the pointer from a Value to its
2169 Type is maintained by PATypeHolder objects.
2174 <!-- ______________________________________________________________________ -->
2175 <div class="doc_subsubsection">
2176 <a name="AbstractTypeUser">The AbstractTypeUser Class</a>
2179 <div class="doc_text">
2182 Some data structures need more to perform more complex updates when types get
2183 resolved. To support this, a class can derive from the AbstractTypeUser class.
2185 allows it to get callbacks when certain types are resolved. To register to get
2186 callbacks for a particular type, the DerivedType::{add/remove}AbstractTypeUser
2187 methods can be called on a type. Note that these methods only work for <i>
2188 abstract</i> types. Concrete types (those that do not include any opaque
2189 objects) can never be refined.
2194 <!-- ======================================================================= -->
2195 <div class="doc_subsection">
2196 <a name="SymbolTable">The <tt>ValueSymbolTable</tt> and
2197 <tt>TypeSymbolTable</tt> classes</a>
2200 <div class="doc_text">
2201 <p>The <tt><a href="http://llvm.org/doxygen/classllvm_1_1ValueSymbolTable.html">
2202 ValueSymbolTable</a></tt> class provides a symbol table that the <a
2203 href="#Function"><tt>Function</tt></a> and <a href="#Module">
2204 <tt>Module</tt></a> classes use for naming value definitions. The symbol table
2205 can provide a name for any <a href="#Value"><tt>Value</tt></a>.
2206 The <tt><a href="http://llvm.org/doxygen/classllvm_1_1TypeSymbolTable.html">
2207 TypeSymbolTable</a></tt> class is used by the <tt>Module</tt> class to store
2208 names for types.</p>
2210 <p>Note that the <tt>SymbolTable</tt> class should not be directly accessed
2211 by most clients. It should only be used when iteration over the symbol table
2212 names themselves are required, which is very special purpose. Note that not
2214 <tt><a href="#Value">Value</a></tt>s have names, and those without names (i.e. they have
2215 an empty name) do not exist in the symbol table.
2218 <p>These symbol tables support iteration over the values/types in the symbol
2219 table with <tt>begin/end/iterator</tt> and supports querying to see if a
2220 specific name is in the symbol table (with <tt>lookup</tt>). The
2221 <tt>ValueSymbolTable</tt> class exposes no public mutator methods, instead,
2222 simply call <tt>setName</tt> on a value, which will autoinsert it into the
2223 appropriate symbol table. For types, use the Module::addTypeName method to
2224 insert entries into the symbol table.</p>
2230 <!-- ======================================================================= -->
2231 <div class="doc_subsection">
2232 <a name="UserLayout">The <tt>User</tt> and owned <tt>Use</tt> classes' memory layout</a>
2235 <div class="doc_text">
2236 <p>The <tt><a href="http://llvm.org/doxygen/classllvm_1_1User.html">
2237 User</a></tt> class provides a basis for expressing the ownership of <tt>User</tt>
2238 towards other <tt><a href="http://llvm.org/doxygen/classllvm_1_1Value.html">
2239 Value</a></tt>s. The <tt><a href="http://llvm.org/doxygen/classllvm_1_1Use.html">
2240 Use</a></tt> helper class is employed to do the bookkeeping and to facilitate <i>O(1)</i>
2241 addition and removal.</p>
2243 <!-- ______________________________________________________________________ -->
2244 <div class="doc_subsubsection">
2245 <a name="Use2User">Interaction and relationship between <tt>User</tt> and <tt>Use</tt> objects</a>
2248 <div class="doc_text">
2250 A subclass of <tt>User</tt> can choose between incorporating its <tt>Use</tt> objects
2251 or refer to them out-of-line by means of a pointer. A mixed variant
2252 (some <tt>Use</tt>s inline others hung off) is impractical and breaks the invariant
2253 that the <tt>Use</tt> objects belonging to the same <tt>User</tt> form a contiguous array.
2258 We have 2 different layouts in the <tt>User</tt> (sub)classes:
2261 The <tt>Use</tt> object(s) are inside (resp. at fixed offset) of the <tt>User</tt>
2262 object and there are a fixed number of them.</p>
2265 The <tt>Use</tt> object(s) are referenced by a pointer to an
2266 array from the <tt>User</tt> object and there may be a variable
2270 As of v2.4 each layout still possesses a direct pointer to the
2271 start of the array of <tt>Use</tt>s. Though not mandatory for layout a),
2272 we stick to this redundancy for the sake of simplicity.
2273 The <tt>User</tt> object also stores the number of <tt>Use</tt> objects it
2274 has. (Theoretically this information can also be calculated
2275 given the scheme presented below.)</p>
2277 Special forms of allocation operators (<tt>operator new</tt>)
2278 enforce the following memory layouts:</p>
2281 <li><p>Layout a) is modelled by prepending the <tt>User</tt> object by the <tt>Use[]</tt> array.</p>
2284 ...---.---.---.---.-------...
2285 | P | P | P | P | User
2286 '''---'---'---'---'-------'''
2289 <li><p>Layout b) is modelled by pointing at the <tt>Use[]</tt> array.</p>
2301 <i>(In the above figures '<tt>P</tt>' stands for the <tt>Use**</tt> that
2302 is stored in each <tt>Use</tt> object in the member <tt>Use::Prev</tt>)</i>
2304 <!-- ______________________________________________________________________ -->
2305 <div class="doc_subsubsection">
2306 <a name="Waymarking">The waymarking algorithm</a>
2309 <div class="doc_text">
2311 Since the <tt>Use</tt> objects are deprived of the direct (back)pointer to
2312 their <tt>User</tt> objects, there must be a fast and exact method to
2313 recover it. This is accomplished by the following scheme:</p>
2316 A bit-encoding in the 2 LSBits (least significant bits) of the <tt>Use::Prev</tt> allows to find the
2317 start of the <tt>User</tt> object:
2319 <li><tt>00</tt> —> binary digit 0</li>
2320 <li><tt>01</tt> —> binary digit 1</li>
2321 <li><tt>10</tt> —> stop and calculate (<tt>s</tt>)</li>
2322 <li><tt>11</tt> —> full stop (<tt>S</tt>)</li>
2325 Given a <tt>Use*</tt>, all we have to do is to walk till we get
2326 a stop and we either have a <tt>User</tt> immediately behind or
2327 we have to walk to the next stop picking up digits
2328 and calculating the offset:</p>
2330 .---.---.---.---.---.---.---.---.---.---.---.---.---.---.---.---.----------------
2331 | 1 | s | 1 | 0 | 1 | 0 | s | 1 | 1 | 0 | s | 1 | 1 | s | 1 | S | User (or User*)
2332 '---'---'---'---'---'---'---'---'---'---'---'---'---'---'---'---'----------------
2333 |+15 |+10 |+6 |+3 |+1
2336 | | |______________________>
2337 | |______________________________________>
2338 |__________________________________________________________>
2341 Only the significant number of bits need to be stored between the
2342 stops, so that the <i>worst case is 20 memory accesses</i> when there are
2343 1000 <tt>Use</tt> objects associated with a <tt>User</tt>.</p>
2345 <!-- ______________________________________________________________________ -->
2346 <div class="doc_subsubsection">
2347 <a name="ReferenceImpl">Reference implementation</a>
2350 <div class="doc_text">
2352 The following literate Haskell fragment demonstrates the concept:</p>
2355 <div class="doc_code">
2357 > import Test.QuickCheck
2359 > digits :: Int -> [Char] -> [Char]
2360 > digits 0 acc = '0' : acc
2361 > digits 1 acc = '1' : acc
2362 > digits n acc = digits (n `div` 2) $ digits (n `mod` 2) acc
2364 > dist :: Int -> [Char] -> [Char]
2367 > dist 1 acc = let r = dist 0 acc in 's' : digits (length r) r
2368 > dist n acc = dist (n - 1) $ dist 1 acc
2370 > takeLast n ss = reverse $ take n $ reverse ss
2372 > test = takeLast 40 $ dist 20 []
2377 Printing <test> gives: <tt>"1s100000s11010s10100s1111s1010s110s11s1S"</tt></p>
2379 The reverse algorithm computes the length of the string just by examining
2380 a certain prefix:</p>
2382 <div class="doc_code">
2384 > pref :: [Char] -> Int
2386 > pref ('s':'1':rest) = decode 2 1 rest
2387 > pref (_:rest) = 1 + pref rest
2389 > decode walk acc ('0':rest) = decode (walk + 1) (acc * 2) rest
2390 > decode walk acc ('1':rest) = decode (walk + 1) (acc * 2 + 1) rest
2391 > decode walk acc _ = walk + acc
2396 Now, as expected, printing <pref test> gives <tt>40</tt>.</p>
2398 We can <i>quickCheck</i> this with following property:</p>
2400 <div class="doc_code">
2402 > testcase = dist 2000 []
2403 > testcaseLength = length testcase
2405 > identityProp n = n > 0 && n <= testcaseLength ==> length arr == pref arr
2406 > where arr = takeLast n testcase
2411 As expected <quickCheck identityProp> gives:</p>
2414 *Main> quickCheck identityProp
2415 OK, passed 100 tests.
2418 Let's be a bit more exhaustive:</p>
2420 <div class="doc_code">
2423 > deepCheck p = check (defaultConfig { configMaxTest = 500 }) p
2428 And here is the result of <deepCheck identityProp>:</p>
2431 *Main> deepCheck identityProp
2432 OK, passed 500 tests.
2435 <!-- ______________________________________________________________________ -->
2436 <div class="doc_subsubsection">
2437 <a name="Tagging">Tagging considerations</a>
2441 To maintain the invariant that the 2 LSBits of each <tt>Use**</tt> in <tt>Use</tt>
2442 never change after being set up, setters of <tt>Use::Prev</tt> must re-tag the
2443 new <tt>Use**</tt> on every modification. Accordingly getters must strip the
2446 For layout b) instead of the <tt>User</tt> we find a pointer (<tt>User*</tt> with LSBit set).
2447 Following this pointer brings us to the <tt>User</tt>. A portable trick ensures
2448 that the first bytes of <tt>User</tt> (if interpreted as a pointer) never has
2449 the LSBit set. (Portability is relying on the fact that all known compilers place the
2450 <tt>vptr</tt> in the first word of the instances.)</p>
2454 <!-- *********************************************************************** -->
2455 <div class="doc_section">
2456 <a name="coreclasses">The Core LLVM Class Hierarchy Reference </a>
2458 <!-- *********************************************************************** -->
2460 <div class="doc_text">
2461 <p><tt>#include "<a href="/doxygen/Type_8h-source.html">llvm/Type.h</a>"</tt>
2462 <br>doxygen info: <a href="/doxygen/classllvm_1_1Type.html">Type Class</a></p>
2464 <p>The Core LLVM classes are the primary means of representing the program
2465 being inspected or transformed. The core LLVM classes are defined in
2466 header files in the <tt>include/llvm/</tt> directory, and implemented in
2467 the <tt>lib/VMCore</tt> directory.</p>
2471 <!-- ======================================================================= -->
2472 <div class="doc_subsection">
2473 <a name="Type">The <tt>Type</tt> class and Derived Types</a>
2476 <div class="doc_text">
2478 <p><tt>Type</tt> is a superclass of all type classes. Every <tt>Value</tt> has
2479 a <tt>Type</tt>. <tt>Type</tt> cannot be instantiated directly but only
2480 through its subclasses. Certain primitive types (<tt>VoidType</tt>,
2481 <tt>LabelType</tt>, <tt>FloatType</tt> and <tt>DoubleType</tt>) have hidden
2482 subclasses. They are hidden because they offer no useful functionality beyond
2483 what the <tt>Type</tt> class offers except to distinguish themselves from
2484 other subclasses of <tt>Type</tt>.</p>
2485 <p>All other types are subclasses of <tt>DerivedType</tt>. Types can be
2486 named, but this is not a requirement. There exists exactly
2487 one instance of a given shape at any one time. This allows type equality to
2488 be performed with address equality of the Type Instance. That is, given two
2489 <tt>Type*</tt> values, the types are identical if the pointers are identical.
2493 <!-- _______________________________________________________________________ -->
2494 <div class="doc_subsubsection">
2495 <a name="m_Type">Important Public Methods</a>
2498 <div class="doc_text">
2501 <li><tt>bool isInteger() const</tt>: Returns true for any integer type.</li>
2503 <li><tt>bool isFloatingPoint()</tt>: Return true if this is one of the two
2504 floating point types.</li>
2506 <li><tt>bool isAbstract()</tt>: Return true if the type is abstract (contains
2507 an OpaqueType anywhere in its definition).</li>
2509 <li><tt>bool isSized()</tt>: Return true if the type has known size. Things
2510 that don't have a size are abstract types, labels and void.</li>
2515 <!-- _______________________________________________________________________ -->
2516 <div class="doc_subsubsection">
2517 <a name="derivedtypes">Important Derived Types</a>
2519 <div class="doc_text">
2521 <dt><tt>IntegerType</tt></dt>
2522 <dd>Subclass of DerivedType that represents integer types of any bit width.
2523 Any bit width between <tt>IntegerType::MIN_INT_BITS</tt> (1) and
2524 <tt>IntegerType::MAX_INT_BITS</tt> (~8 million) can be represented.
2526 <li><tt>static const IntegerType* get(unsigned NumBits)</tt>: get an integer
2527 type of a specific bit width.</li>
2528 <li><tt>unsigned getBitWidth() const</tt>: Get the bit width of an integer
2532 <dt><tt>SequentialType</tt></dt>
2533 <dd>This is subclassed by ArrayType and PointerType
2535 <li><tt>const Type * getElementType() const</tt>: Returns the type of each
2536 of the elements in the sequential type. </li>
2539 <dt><tt>ArrayType</tt></dt>
2540 <dd>This is a subclass of SequentialType and defines the interface for array
2543 <li><tt>unsigned getNumElements() const</tt>: Returns the number of
2544 elements in the array. </li>
2547 <dt><tt>PointerType</tt></dt>
2548 <dd>Subclass of SequentialType for pointer types.</dd>
2549 <dt><tt>VectorType</tt></dt>
2550 <dd>Subclass of SequentialType for vector types. A
2551 vector type is similar to an ArrayType but is distinguished because it is
2552 a first class type wherease ArrayType is not. Vector types are used for
2553 vector operations and are usually small vectors of of an integer or floating
2555 <dt><tt>StructType</tt></dt>
2556 <dd>Subclass of DerivedTypes for struct types.</dd>
2557 <dt><tt><a name="FunctionType">FunctionType</a></tt></dt>
2558 <dd>Subclass of DerivedTypes for function types.
2560 <li><tt>bool isVarArg() const</tt>: Returns true if its a vararg
2562 <li><tt> const Type * getReturnType() const</tt>: Returns the
2563 return type of the function.</li>
2564 <li><tt>const Type * getParamType (unsigned i)</tt>: Returns
2565 the type of the ith parameter.</li>
2566 <li><tt> const unsigned getNumParams() const</tt>: Returns the
2567 number of formal parameters.</li>
2570 <dt><tt>OpaqueType</tt></dt>
2571 <dd>Sublcass of DerivedType for abstract types. This class
2572 defines no content and is used as a placeholder for some other type. Note
2573 that OpaqueType is used (temporarily) during type resolution for forward
2574 references of types. Once the referenced type is resolved, the OpaqueType
2575 is replaced with the actual type. OpaqueType can also be used for data
2576 abstraction. At link time opaque types can be resolved to actual types
2577 of the same name.</dd>
2583 <!-- ======================================================================= -->
2584 <div class="doc_subsection">
2585 <a name="Module">The <tt>Module</tt> class</a>
2588 <div class="doc_text">
2591 href="/doxygen/Module_8h-source.html">llvm/Module.h</a>"</tt><br> doxygen info:
2592 <a href="/doxygen/classllvm_1_1Module.html">Module Class</a></p>
2594 <p>The <tt>Module</tt> class represents the top level structure present in LLVM
2595 programs. An LLVM module is effectively either a translation unit of the
2596 original program or a combination of several translation units merged by the
2597 linker. The <tt>Module</tt> class keeps track of a list of <a
2598 href="#Function"><tt>Function</tt></a>s, a list of <a
2599 href="#GlobalVariable"><tt>GlobalVariable</tt></a>s, and a <a
2600 href="#SymbolTable"><tt>SymbolTable</tt></a>. Additionally, it contains a few
2601 helpful member functions that try to make common operations easy.</p>
2605 <!-- _______________________________________________________________________ -->
2606 <div class="doc_subsubsection">
2607 <a name="m_Module">Important Public Members of the <tt>Module</tt> class</a>
2610 <div class="doc_text">
2613 <li><tt>Module::Module(std::string name = "")</tt></li>
2616 <p>Constructing a <a href="#Module">Module</a> is easy. You can optionally
2617 provide a name for it (probably based on the name of the translation unit).</p>
2620 <li><tt>Module::iterator</tt> - Typedef for function list iterator<br>
2621 <tt>Module::const_iterator</tt> - Typedef for const_iterator.<br>
2623 <tt>begin()</tt>, <tt>end()</tt>
2624 <tt>size()</tt>, <tt>empty()</tt>
2626 <p>These are forwarding methods that make it easy to access the contents of
2627 a <tt>Module</tt> object's <a href="#Function"><tt>Function</tt></a>
2630 <li><tt>Module::FunctionListType &getFunctionList()</tt>
2632 <p> Returns the list of <a href="#Function"><tt>Function</tt></a>s. This is
2633 necessary to use when you need to update the list or perform a complex
2634 action that doesn't have a forwarding method.</p>
2636 <p><!-- Global Variable --></p></li>
2642 <li><tt>Module::global_iterator</tt> - Typedef for global variable list iterator<br>
2644 <tt>Module::const_global_iterator</tt> - Typedef for const_iterator.<br>
2646 <tt>global_begin()</tt>, <tt>global_end()</tt>
2647 <tt>global_size()</tt>, <tt>global_empty()</tt>
2649 <p> These are forwarding methods that make it easy to access the contents of
2650 a <tt>Module</tt> object's <a
2651 href="#GlobalVariable"><tt>GlobalVariable</tt></a> list.</p></li>
2653 <li><tt>Module::GlobalListType &getGlobalList()</tt>
2655 <p>Returns the list of <a
2656 href="#GlobalVariable"><tt>GlobalVariable</tt></a>s. This is necessary to
2657 use when you need to update the list or perform a complex action that
2658 doesn't have a forwarding method.</p>
2660 <p><!-- Symbol table stuff --> </p></li>
2666 <li><tt><a href="#SymbolTable">SymbolTable</a> *getSymbolTable()</tt>
2668 <p>Return a reference to the <a href="#SymbolTable"><tt>SymbolTable</tt></a>
2669 for this <tt>Module</tt>.</p>
2671 <p><!-- Convenience methods --></p></li>
2677 <li><tt><a href="#Function">Function</a> *getFunction(const std::string
2678 &Name, const <a href="#FunctionType">FunctionType</a> *Ty)</tt>
2680 <p>Look up the specified function in the <tt>Module</tt> <a
2681 href="#SymbolTable"><tt>SymbolTable</tt></a>. If it does not exist, return
2682 <tt>null</tt>.</p></li>
2684 <li><tt><a href="#Function">Function</a> *getOrInsertFunction(const
2685 std::string &Name, const <a href="#FunctionType">FunctionType</a> *T)</tt>
2687 <p>Look up the specified function in the <tt>Module</tt> <a
2688 href="#SymbolTable"><tt>SymbolTable</tt></a>. If it does not exist, add an
2689 external declaration for the function and return it.</p></li>
2691 <li><tt>std::string getTypeName(const <a href="#Type">Type</a> *Ty)</tt>
2693 <p>If there is at least one entry in the <a
2694 href="#SymbolTable"><tt>SymbolTable</tt></a> for the specified <a
2695 href="#Type"><tt>Type</tt></a>, return it. Otherwise return the empty
2698 <li><tt>bool addTypeName(const std::string &Name, const <a
2699 href="#Type">Type</a> *Ty)</tt>
2701 <p>Insert an entry in the <a href="#SymbolTable"><tt>SymbolTable</tt></a>
2702 mapping <tt>Name</tt> to <tt>Ty</tt>. If there is already an entry for this
2703 name, true is returned and the <a
2704 href="#SymbolTable"><tt>SymbolTable</tt></a> is not modified.</p></li>
2710 <!-- ======================================================================= -->
2711 <div class="doc_subsection">
2712 <a name="Value">The <tt>Value</tt> class</a>
2715 <div class="doc_text">
2717 <p><tt>#include "<a href="/doxygen/Value_8h-source.html">llvm/Value.h</a>"</tt>
2719 doxygen info: <a href="/doxygen/classllvm_1_1Value.html">Value Class</a></p>
2721 <p>The <tt>Value</tt> class is the most important class in the LLVM Source
2722 base. It represents a typed value that may be used (among other things) as an
2723 operand to an instruction. There are many different types of <tt>Value</tt>s,
2724 such as <a href="#Constant"><tt>Constant</tt></a>s,<a
2725 href="#Argument"><tt>Argument</tt></a>s. Even <a
2726 href="#Instruction"><tt>Instruction</tt></a>s and <a
2727 href="#Function"><tt>Function</tt></a>s are <tt>Value</tt>s.</p>
2729 <p>A particular <tt>Value</tt> may be used many times in the LLVM representation
2730 for a program. For example, an incoming argument to a function (represented
2731 with an instance of the <a href="#Argument">Argument</a> class) is "used" by
2732 every instruction in the function that references the argument. To keep track
2733 of this relationship, the <tt>Value</tt> class keeps a list of all of the <a
2734 href="#User"><tt>User</tt></a>s that is using it (the <a
2735 href="#User"><tt>User</tt></a> class is a base class for all nodes in the LLVM
2736 graph that can refer to <tt>Value</tt>s). This use list is how LLVM represents
2737 def-use information in the program, and is accessible through the <tt>use_</tt>*
2738 methods, shown below.</p>
2740 <p>Because LLVM is a typed representation, every LLVM <tt>Value</tt> is typed,
2741 and this <a href="#Type">Type</a> is available through the <tt>getType()</tt>
2742 method. In addition, all LLVM values can be named. The "name" of the
2743 <tt>Value</tt> is a symbolic string printed in the LLVM code:</p>
2745 <div class="doc_code">
2747 %<b>foo</b> = add i32 1, 2
2751 <p><a name="nameWarning">The name of this instruction is "foo".</a> <b>NOTE</b>
2752 that the name of any value may be missing (an empty string), so names should
2753 <b>ONLY</b> be used for debugging (making the source code easier to read,
2754 debugging printouts), they should not be used to keep track of values or map
2755 between them. For this purpose, use a <tt>std::map</tt> of pointers to the
2756 <tt>Value</tt> itself instead.</p>
2758 <p>One important aspect of LLVM is that there is no distinction between an SSA
2759 variable and the operation that produces it. Because of this, any reference to
2760 the value produced by an instruction (or the value available as an incoming
2761 argument, for example) is represented as a direct pointer to the instance of
2763 represents this value. Although this may take some getting used to, it
2764 simplifies the representation and makes it easier to manipulate.</p>
2768 <!-- _______________________________________________________________________ -->
2769 <div class="doc_subsubsection">
2770 <a name="m_Value">Important Public Members of the <tt>Value</tt> class</a>
2773 <div class="doc_text">
2776 <li><tt>Value::use_iterator</tt> - Typedef for iterator over the
2778 <tt>Value::use_const_iterator</tt> - Typedef for const_iterator over
2780 <tt>unsigned use_size()</tt> - Returns the number of users of the
2782 <tt>bool use_empty()</tt> - Returns true if there are no users.<br>
2783 <tt>use_iterator use_begin()</tt> - Get an iterator to the start of
2785 <tt>use_iterator use_end()</tt> - Get an iterator to the end of the
2787 <tt><a href="#User">User</a> *use_back()</tt> - Returns the last
2788 element in the list.
2789 <p> These methods are the interface to access the def-use
2790 information in LLVM. As with all other iterators in LLVM, the naming
2791 conventions follow the conventions defined by the <a href="#stl">STL</a>.</p>
2793 <li><tt><a href="#Type">Type</a> *getType() const</tt>
2794 <p>This method returns the Type of the Value.</p>
2796 <li><tt>bool hasName() const</tt><br>
2797 <tt>std::string getName() const</tt><br>
2798 <tt>void setName(const std::string &Name)</tt>
2799 <p> This family of methods is used to access and assign a name to a <tt>Value</tt>,
2800 be aware of the <a href="#nameWarning">precaution above</a>.</p>
2802 <li><tt>void replaceAllUsesWith(Value *V)</tt>
2804 <p>This method traverses the use list of a <tt>Value</tt> changing all <a
2805 href="#User"><tt>User</tt>s</a> of the current value to refer to
2806 "<tt>V</tt>" instead. For example, if you detect that an instruction always
2807 produces a constant value (for example through constant folding), you can
2808 replace all uses of the instruction with the constant like this:</p>
2810 <div class="doc_code">
2812 Inst->replaceAllUsesWith(ConstVal);
2820 <!-- ======================================================================= -->
2821 <div class="doc_subsection">
2822 <a name="User">The <tt>User</tt> class</a>
2825 <div class="doc_text">
2828 <tt>#include "<a href="/doxygen/User_8h-source.html">llvm/User.h</a>"</tt><br>
2829 doxygen info: <a href="/doxygen/classllvm_1_1User.html">User Class</a><br>
2830 Superclass: <a href="#Value"><tt>Value</tt></a></p>
2832 <p>The <tt>User</tt> class is the common base class of all LLVM nodes that may
2833 refer to <a href="#Value"><tt>Value</tt></a>s. It exposes a list of "Operands"
2834 that are all of the <a href="#Value"><tt>Value</tt></a>s that the User is
2835 referring to. The <tt>User</tt> class itself is a subclass of
2838 <p>The operands of a <tt>User</tt> point directly to the LLVM <a
2839 href="#Value"><tt>Value</tt></a> that it refers to. Because LLVM uses Static
2840 Single Assignment (SSA) form, there can only be one definition referred to,
2841 allowing this direct connection. This connection provides the use-def
2842 information in LLVM.</p>
2846 <!-- _______________________________________________________________________ -->
2847 <div class="doc_subsubsection">
2848 <a name="m_User">Important Public Members of the <tt>User</tt> class</a>
2851 <div class="doc_text">
2853 <p>The <tt>User</tt> class exposes the operand list in two ways: through
2854 an index access interface and through an iterator based interface.</p>
2857 <li><tt>Value *getOperand(unsigned i)</tt><br>
2858 <tt>unsigned getNumOperands()</tt>
2859 <p> These two methods expose the operands of the <tt>User</tt> in a
2860 convenient form for direct access.</p></li>
2862 <li><tt>User::op_iterator</tt> - Typedef for iterator over the operand
2864 <tt>op_iterator op_begin()</tt> - Get an iterator to the start of
2865 the operand list.<br>
2866 <tt>op_iterator op_end()</tt> - Get an iterator to the end of the
2868 <p> Together, these methods make up the iterator based interface to
2869 the operands of a <tt>User</tt>.</p></li>
2874 <!-- ======================================================================= -->
2875 <div class="doc_subsection">
2876 <a name="Instruction">The <tt>Instruction</tt> class</a>
2879 <div class="doc_text">
2881 <p><tt>#include "</tt><tt><a
2882 href="/doxygen/Instruction_8h-source.html">llvm/Instruction.h</a>"</tt><br>
2883 doxygen info: <a href="/doxygen/classllvm_1_1Instruction.html">Instruction Class</a><br>
2884 Superclasses: <a href="#User"><tt>User</tt></a>, <a
2885 href="#Value"><tt>Value</tt></a></p>
2887 <p>The <tt>Instruction</tt> class is the common base class for all LLVM
2888 instructions. It provides only a few methods, but is a very commonly used
2889 class. The primary data tracked by the <tt>Instruction</tt> class itself is the
2890 opcode (instruction type) and the parent <a
2891 href="#BasicBlock"><tt>BasicBlock</tt></a> the <tt>Instruction</tt> is embedded
2892 into. To represent a specific type of instruction, one of many subclasses of
2893 <tt>Instruction</tt> are used.</p>
2895 <p> Because the <tt>Instruction</tt> class subclasses the <a
2896 href="#User"><tt>User</tt></a> class, its operands can be accessed in the same
2897 way as for other <a href="#User"><tt>User</tt></a>s (with the
2898 <tt>getOperand()</tt>/<tt>getNumOperands()</tt> and
2899 <tt>op_begin()</tt>/<tt>op_end()</tt> methods).</p> <p> An important file for
2900 the <tt>Instruction</tt> class is the <tt>llvm/Instruction.def</tt> file. This
2901 file contains some meta-data about the various different types of instructions
2902 in LLVM. It describes the enum values that are used as opcodes (for example
2903 <tt>Instruction::Add</tt> and <tt>Instruction::ICmp</tt>), as well as the
2904 concrete sub-classes of <tt>Instruction</tt> that implement the instruction (for
2905 example <tt><a href="#BinaryOperator">BinaryOperator</a></tt> and <tt><a
2906 href="#CmpInst">CmpInst</a></tt>). Unfortunately, the use of macros in
2907 this file confuses doxygen, so these enum values don't show up correctly in the
2908 <a href="/doxygen/classllvm_1_1Instruction.html">doxygen output</a>.</p>
2912 <!-- _______________________________________________________________________ -->
2913 <div class="doc_subsubsection">
2914 <a name="s_Instruction">Important Subclasses of the <tt>Instruction</tt>
2917 <div class="doc_text">
2919 <li><tt><a name="BinaryOperator">BinaryOperator</a></tt>
2920 <p>This subclasses represents all two operand instructions whose operands
2921 must be the same type, except for the comparison instructions.</p></li>
2922 <li><tt><a name="CastInst">CastInst</a></tt>
2923 <p>This subclass is the parent of the 12 casting instructions. It provides
2924 common operations on cast instructions.</p>
2925 <li><tt><a name="CmpInst">CmpInst</a></tt>
2926 <p>This subclass respresents the two comparison instructions,
2927 <a href="LangRef.html#i_icmp">ICmpInst</a> (integer opreands), and
2928 <a href="LangRef.html#i_fcmp">FCmpInst</a> (floating point operands).</p>
2929 <li><tt><a name="TerminatorInst">TerminatorInst</a></tt>
2930 <p>This subclass is the parent of all terminator instructions (those which
2931 can terminate a block).</p>
2935 <!-- _______________________________________________________________________ -->
2936 <div class="doc_subsubsection">
2937 <a name="m_Instruction">Important Public Members of the <tt>Instruction</tt>
2941 <div class="doc_text">
2944 <li><tt><a href="#BasicBlock">BasicBlock</a> *getParent()</tt>
2945 <p>Returns the <a href="#BasicBlock"><tt>BasicBlock</tt></a> that
2946 this <tt>Instruction</tt> is embedded into.</p></li>
2947 <li><tt>bool mayWriteToMemory()</tt>
2948 <p>Returns true if the instruction writes to memory, i.e. it is a
2949 <tt>call</tt>,<tt>free</tt>,<tt>invoke</tt>, or <tt>store</tt>.</p></li>
2950 <li><tt>unsigned getOpcode()</tt>
2951 <p>Returns the opcode for the <tt>Instruction</tt>.</p></li>
2952 <li><tt><a href="#Instruction">Instruction</a> *clone() const</tt>
2953 <p>Returns another instance of the specified instruction, identical
2954 in all ways to the original except that the instruction has no parent
2955 (ie it's not embedded into a <a href="#BasicBlock"><tt>BasicBlock</tt></a>),
2956 and it has no name</p></li>
2961 <!-- ======================================================================= -->
2962 <div class="doc_subsection">
2963 <a name="Constant">The <tt>Constant</tt> class and subclasses</a>
2966 <div class="doc_text">
2968 <p>Constant represents a base class for different types of constants. It
2969 is subclassed by ConstantInt, ConstantArray, etc. for representing
2970 the various types of Constants. <a href="#GlobalValue">GlobalValue</a> is also
2971 a subclass, which represents the address of a global variable or function.
2976 <!-- _______________________________________________________________________ -->
2977 <div class="doc_subsubsection">Important Subclasses of Constant </div>
2978 <div class="doc_text">
2980 <li>ConstantInt : This subclass of Constant represents an integer constant of
2983 <li><tt>const APInt& getValue() const</tt>: Returns the underlying
2984 value of this constant, an APInt value.</li>
2985 <li><tt>int64_t getSExtValue() const</tt>: Converts the underlying APInt
2986 value to an int64_t via sign extension. If the value (not the bit width)
2987 of the APInt is too large to fit in an int64_t, an assertion will result.
2988 For this reason, use of this method is discouraged.</li>
2989 <li><tt>uint64_t getZExtValue() const</tt>: Converts the underlying APInt
2990 value to a uint64_t via zero extension. IF the value (not the bit width)
2991 of the APInt is too large to fit in a uint64_t, an assertion will result.
2992 For this reason, use of this method is discouraged.</li>
2993 <li><tt>static ConstantInt* get(const APInt& Val)</tt>: Returns the
2994 ConstantInt object that represents the value provided by <tt>Val</tt>.
2995 The type is implied as the IntegerType that corresponds to the bit width
2996 of <tt>Val</tt>.</li>
2997 <li><tt>static ConstantInt* get(const Type *Ty, uint64_t Val)</tt>:
2998 Returns the ConstantInt object that represents the value provided by
2999 <tt>Val</tt> for integer type <tt>Ty</tt>.</li>
3002 <li>ConstantFP : This class represents a floating point constant.
3004 <li><tt>double getValue() const</tt>: Returns the underlying value of
3005 this constant. </li>
3008 <li>ConstantArray : This represents a constant array.
3010 <li><tt>const std::vector<Use> &getValues() const</tt>: Returns
3011 a vector of component constants that makeup this array. </li>
3014 <li>ConstantStruct : This represents a constant struct.
3016 <li><tt>const std::vector<Use> &getValues() const</tt>: Returns
3017 a vector of component constants that makeup this array. </li>
3020 <li>GlobalValue : This represents either a global variable or a function. In
3021 either case, the value is a constant fixed address (after linking).
3027 <!-- ======================================================================= -->
3028 <div class="doc_subsection">
3029 <a name="GlobalValue">The <tt>GlobalValue</tt> class</a>
3032 <div class="doc_text">
3035 href="/doxygen/GlobalValue_8h-source.html">llvm/GlobalValue.h</a>"</tt><br>
3036 doxygen info: <a href="/doxygen/classllvm_1_1GlobalValue.html">GlobalValue
3038 Superclasses: <a href="#Constant"><tt>Constant</tt></a>,
3039 <a href="#User"><tt>User</tt></a>, <a href="#Value"><tt>Value</tt></a></p>
3041 <p>Global values (<a href="#GlobalVariable"><tt>GlobalVariable</tt></a>s or <a
3042 href="#Function"><tt>Function</tt></a>s) are the only LLVM values that are
3043 visible in the bodies of all <a href="#Function"><tt>Function</tt></a>s.
3044 Because they are visible at global scope, they are also subject to linking with
3045 other globals defined in different translation units. To control the linking
3046 process, <tt>GlobalValue</tt>s know their linkage rules. Specifically,
3047 <tt>GlobalValue</tt>s know whether they have internal or external linkage, as
3048 defined by the <tt>LinkageTypes</tt> enumeration.</p>
3050 <p>If a <tt>GlobalValue</tt> has internal linkage (equivalent to being
3051 <tt>static</tt> in C), it is not visible to code outside the current translation
3052 unit, and does not participate in linking. If it has external linkage, it is
3053 visible to external code, and does participate in linking. In addition to
3054 linkage information, <tt>GlobalValue</tt>s keep track of which <a
3055 href="#Module"><tt>Module</tt></a> they are currently part of.</p>
3057 <p>Because <tt>GlobalValue</tt>s are memory objects, they are always referred to
3058 by their <b>address</b>. As such, the <a href="#Type"><tt>Type</tt></a> of a
3059 global is always a pointer to its contents. It is important to remember this
3060 when using the <tt>GetElementPtrInst</tt> instruction because this pointer must
3061 be dereferenced first. For example, if you have a <tt>GlobalVariable</tt> (a
3062 subclass of <tt>GlobalValue)</tt> that is an array of 24 ints, type <tt>[24 x
3063 i32]</tt>, then the <tt>GlobalVariable</tt> is a pointer to that array. Although
3064 the address of the first element of this array and the value of the
3065 <tt>GlobalVariable</tt> are the same, they have different types. The
3066 <tt>GlobalVariable</tt>'s type is <tt>[24 x i32]</tt>. The first element's type
3067 is <tt>i32.</tt> Because of this, accessing a global value requires you to
3068 dereference the pointer with <tt>GetElementPtrInst</tt> first, then its elements
3069 can be accessed. This is explained in the <a href="LangRef.html#globalvars">LLVM
3070 Language Reference Manual</a>.</p>
3074 <!-- _______________________________________________________________________ -->
3075 <div class="doc_subsubsection">
3076 <a name="m_GlobalValue">Important Public Members of the <tt>GlobalValue</tt>
3080 <div class="doc_text">
3083 <li><tt>bool hasInternalLinkage() const</tt><br>
3084 <tt>bool hasExternalLinkage() const</tt><br>
3085 <tt>void setInternalLinkage(bool HasInternalLinkage)</tt>
3086 <p> These methods manipulate the linkage characteristics of the <tt>GlobalValue</tt>.</p>
3089 <li><tt><a href="#Module">Module</a> *getParent()</tt>
3090 <p> This returns the <a href="#Module"><tt>Module</tt></a> that the
3091 GlobalValue is currently embedded into.</p></li>
3096 <!-- ======================================================================= -->
3097 <div class="doc_subsection">
3098 <a name="Function">The <tt>Function</tt> class</a>
3101 <div class="doc_text">
3104 href="/doxygen/Function_8h-source.html">llvm/Function.h</a>"</tt><br> doxygen
3105 info: <a href="/doxygen/classllvm_1_1Function.html">Function Class</a><br>
3106 Superclasses: <a href="#GlobalValue"><tt>GlobalValue</tt></a>,
3107 <a href="#Constant"><tt>Constant</tt></a>,
3108 <a href="#User"><tt>User</tt></a>,
3109 <a href="#Value"><tt>Value</tt></a></p>
3111 <p>The <tt>Function</tt> class represents a single procedure in LLVM. It is
3112 actually one of the more complex classes in the LLVM heirarchy because it must
3113 keep track of a large amount of data. The <tt>Function</tt> class keeps track
3114 of a list of <a href="#BasicBlock"><tt>BasicBlock</tt></a>s, a list of formal
3115 <a href="#Argument"><tt>Argument</tt></a>s, and a
3116 <a href="#SymbolTable"><tt>SymbolTable</tt></a>.</p>
3118 <p>The list of <a href="#BasicBlock"><tt>BasicBlock</tt></a>s is the most
3119 commonly used part of <tt>Function</tt> objects. The list imposes an implicit
3120 ordering of the blocks in the function, which indicate how the code will be
3121 layed out by the backend. Additionally, the first <a
3122 href="#BasicBlock"><tt>BasicBlock</tt></a> is the implicit entry node for the
3123 <tt>Function</tt>. It is not legal in LLVM to explicitly branch to this initial
3124 block. There are no implicit exit nodes, and in fact there may be multiple exit
3125 nodes from a single <tt>Function</tt>. If the <a
3126 href="#BasicBlock"><tt>BasicBlock</tt></a> list is empty, this indicates that
3127 the <tt>Function</tt> is actually a function declaration: the actual body of the
3128 function hasn't been linked in yet.</p>
3130 <p>In addition to a list of <a href="#BasicBlock"><tt>BasicBlock</tt></a>s, the
3131 <tt>Function</tt> class also keeps track of the list of formal <a
3132 href="#Argument"><tt>Argument</tt></a>s that the function receives. This
3133 container manages the lifetime of the <a href="#Argument"><tt>Argument</tt></a>
3134 nodes, just like the <a href="#BasicBlock"><tt>BasicBlock</tt></a> list does for
3135 the <a href="#BasicBlock"><tt>BasicBlock</tt></a>s.</p>
3137 <p>The <a href="#SymbolTable"><tt>SymbolTable</tt></a> is a very rarely used
3138 LLVM feature that is only used when you have to look up a value by name. Aside
3139 from that, the <a href="#SymbolTable"><tt>SymbolTable</tt></a> is used
3140 internally to make sure that there are not conflicts between the names of <a
3141 href="#Instruction"><tt>Instruction</tt></a>s, <a
3142 href="#BasicBlock"><tt>BasicBlock</tt></a>s, or <a
3143 href="#Argument"><tt>Argument</tt></a>s in the function body.</p>
3145 <p>Note that <tt>Function</tt> is a <a href="#GlobalValue">GlobalValue</a>
3146 and therefore also a <a href="#Constant">Constant</a>. The value of the function
3147 is its address (after linking) which is guaranteed to be constant.</p>
3150 <!-- _______________________________________________________________________ -->
3151 <div class="doc_subsubsection">
3152 <a name="m_Function">Important Public Members of the <tt>Function</tt>
3156 <div class="doc_text">
3159 <li><tt>Function(const </tt><tt><a href="#FunctionType">FunctionType</a>
3160 *Ty, LinkageTypes Linkage, const std::string &N = "", Module* Parent = 0)</tt>
3162 <p>Constructor used when you need to create new <tt>Function</tt>s to add
3163 the the program. The constructor must specify the type of the function to
3164 create and what type of linkage the function should have. The <a
3165 href="#FunctionType"><tt>FunctionType</tt></a> argument
3166 specifies the formal arguments and return value for the function. The same
3167 <a href="#FunctionType"><tt>FunctionType</tt></a> value can be used to
3168 create multiple functions. The <tt>Parent</tt> argument specifies the Module
3169 in which the function is defined. If this argument is provided, the function
3170 will automatically be inserted into that module's list of
3173 <li><tt>bool isDeclaration()</tt>
3175 <p>Return whether or not the <tt>Function</tt> has a body defined. If the
3176 function is "external", it does not have a body, and thus must be resolved
3177 by linking with a function defined in a different translation unit.</p></li>
3179 <li><tt>Function::iterator</tt> - Typedef for basic block list iterator<br>
3180 <tt>Function::const_iterator</tt> - Typedef for const_iterator.<br>
3182 <tt>begin()</tt>, <tt>end()</tt>
3183 <tt>size()</tt>, <tt>empty()</tt>
3185 <p>These are forwarding methods that make it easy to access the contents of
3186 a <tt>Function</tt> object's <a href="#BasicBlock"><tt>BasicBlock</tt></a>
3189 <li><tt>Function::BasicBlockListType &getBasicBlockList()</tt>
3191 <p>Returns the list of <a href="#BasicBlock"><tt>BasicBlock</tt></a>s. This
3192 is necessary to use when you need to update the list or perform a complex
3193 action that doesn't have a forwarding method.</p></li>
3195 <li><tt>Function::arg_iterator</tt> - Typedef for the argument list
3197 <tt>Function::const_arg_iterator</tt> - Typedef for const_iterator.<br>
3199 <tt>arg_begin()</tt>, <tt>arg_end()</tt>
3200 <tt>arg_size()</tt>, <tt>arg_empty()</tt>
3202 <p>These are forwarding methods that make it easy to access the contents of
3203 a <tt>Function</tt> object's <a href="#Argument"><tt>Argument</tt></a>
3206 <li><tt>Function::ArgumentListType &getArgumentList()</tt>
3208 <p>Returns the list of <a href="#Argument"><tt>Argument</tt></a>s. This is
3209 necessary to use when you need to update the list or perform a complex
3210 action that doesn't have a forwarding method.</p></li>
3212 <li><tt><a href="#BasicBlock">BasicBlock</a> &getEntryBlock()</tt>
3214 <p>Returns the entry <a href="#BasicBlock"><tt>BasicBlock</tt></a> for the
3215 function. Because the entry block for the function is always the first
3216 block, this returns the first block of the <tt>Function</tt>.</p></li>
3218 <li><tt><a href="#Type">Type</a> *getReturnType()</tt><br>
3219 <tt><a href="#FunctionType">FunctionType</a> *getFunctionType()</tt>
3221 <p>This traverses the <a href="#Type"><tt>Type</tt></a> of the
3222 <tt>Function</tt> and returns the return type of the function, or the <a
3223 href="#FunctionType"><tt>FunctionType</tt></a> of the actual
3226 <li><tt><a href="#SymbolTable">SymbolTable</a> *getSymbolTable()</tt>
3228 <p> Return a pointer to the <a href="#SymbolTable"><tt>SymbolTable</tt></a>
3229 for this <tt>Function</tt>.</p></li>
3234 <!-- ======================================================================= -->
3235 <div class="doc_subsection">
3236 <a name="GlobalVariable">The <tt>GlobalVariable</tt> class</a>
3239 <div class="doc_text">
3242 href="/doxygen/GlobalVariable_8h-source.html">llvm/GlobalVariable.h</a>"</tt>
3244 doxygen info: <a href="/doxygen/classllvm_1_1GlobalVariable.html">GlobalVariable
3246 Superclasses: <a href="#GlobalValue"><tt>GlobalValue</tt></a>,
3247 <a href="#Constant"><tt>Constant</tt></a>,
3248 <a href="#User"><tt>User</tt></a>,
3249 <a href="#Value"><tt>Value</tt></a></p>
3251 <p>Global variables are represented with the (suprise suprise)
3252 <tt>GlobalVariable</tt> class. Like functions, <tt>GlobalVariable</tt>s are also
3253 subclasses of <a href="#GlobalValue"><tt>GlobalValue</tt></a>, and as such are
3254 always referenced by their address (global values must live in memory, so their
3255 "name" refers to their constant address). See
3256 <a href="#GlobalValue"><tt>GlobalValue</tt></a> for more on this. Global
3257 variables may have an initial value (which must be a
3258 <a href="#Constant"><tt>Constant</tt></a>), and if they have an initializer,
3259 they may be marked as "constant" themselves (indicating that their contents
3260 never change at runtime).</p>
3263 <!-- _______________________________________________________________________ -->
3264 <div class="doc_subsubsection">
3265 <a name="m_GlobalVariable">Important Public Members of the
3266 <tt>GlobalVariable</tt> class</a>
3269 <div class="doc_text">
3272 <li><tt>GlobalVariable(const </tt><tt><a href="#Type">Type</a> *Ty, bool
3273 isConstant, LinkageTypes& Linkage, <a href="#Constant">Constant</a>
3274 *Initializer = 0, const std::string &Name = "", Module* Parent = 0)</tt>
3276 <p>Create a new global variable of the specified type. If
3277 <tt>isConstant</tt> is true then the global variable will be marked as
3278 unchanging for the program. The Linkage parameter specifies the type of
3279 linkage (internal, external, weak, linkonce, appending) for the variable. If
3280 the linkage is InternalLinkage, WeakLinkage, or LinkOnceLinkage, then
3281 the resultant global variable will have internal linkage. AppendingLinkage
3282 concatenates together all instances (in different translation units) of the
3283 variable into a single variable but is only applicable to arrays. See
3284 the <a href="LangRef.html#modulestructure">LLVM Language Reference</a> for
3285 further details on linkage types. Optionally an initializer, a name, and the
3286 module to put the variable into may be specified for the global variable as
3289 <li><tt>bool isConstant() const</tt>
3291 <p>Returns true if this is a global variable that is known not to
3292 be modified at runtime.</p></li>
3294 <li><tt>bool hasInitializer()</tt>
3296 <p>Returns true if this <tt>GlobalVariable</tt> has an intializer.</p></li>
3298 <li><tt><a href="#Constant">Constant</a> *getInitializer()</tt>
3300 <p>Returns the intial value for a <tt>GlobalVariable</tt>. It is not legal
3301 to call this method if there is no initializer.</p></li>
3307 <!-- ======================================================================= -->
3308 <div class="doc_subsection">
3309 <a name="BasicBlock">The <tt>BasicBlock</tt> class</a>
3312 <div class="doc_text">
3315 href="/doxygen/BasicBlock_8h-source.html">llvm/BasicBlock.h</a>"</tt><br>
3316 doxygen info: <a href="/doxygen/structllvm_1_1BasicBlock.html">BasicBlock
3318 Superclass: <a href="#Value"><tt>Value</tt></a></p>
3320 <p>This class represents a single entry multiple exit section of the code,
3321 commonly known as a basic block by the compiler community. The
3322 <tt>BasicBlock</tt> class maintains a list of <a
3323 href="#Instruction"><tt>Instruction</tt></a>s, which form the body of the block.
3324 Matching the language definition, the last element of this list of instructions
3325 is always a terminator instruction (a subclass of the <a
3326 href="#TerminatorInst"><tt>TerminatorInst</tt></a> class).</p>
3328 <p>In addition to tracking the list of instructions that make up the block, the
3329 <tt>BasicBlock</tt> class also keeps track of the <a
3330 href="#Function"><tt>Function</tt></a> that it is embedded into.</p>
3332 <p>Note that <tt>BasicBlock</tt>s themselves are <a
3333 href="#Value"><tt>Value</tt></a>s, because they are referenced by instructions
3334 like branches and can go in the switch tables. <tt>BasicBlock</tt>s have type
3339 <!-- _______________________________________________________________________ -->
3340 <div class="doc_subsubsection">
3341 <a name="m_BasicBlock">Important Public Members of the <tt>BasicBlock</tt>
3345 <div class="doc_text">
3348 <li><tt>BasicBlock(const std::string &Name = "", </tt><tt><a
3349 href="#Function">Function</a> *Parent = 0)</tt>
3351 <p>The <tt>BasicBlock</tt> constructor is used to create new basic blocks for
3352 insertion into a function. The constructor optionally takes a name for the new
3353 block, and a <a href="#Function"><tt>Function</tt></a> to insert it into. If
3354 the <tt>Parent</tt> parameter is specified, the new <tt>BasicBlock</tt> is
3355 automatically inserted at the end of the specified <a
3356 href="#Function"><tt>Function</tt></a>, if not specified, the BasicBlock must be
3357 manually inserted into the <a href="#Function"><tt>Function</tt></a>.</p></li>
3359 <li><tt>BasicBlock::iterator</tt> - Typedef for instruction list iterator<br>
3360 <tt>BasicBlock::const_iterator</tt> - Typedef for const_iterator.<br>
3361 <tt>begin()</tt>, <tt>end()</tt>, <tt>front()</tt>, <tt>back()</tt>,
3362 <tt>size()</tt>, <tt>empty()</tt>
3363 STL-style functions for accessing the instruction list.
3365 <p>These methods and typedefs are forwarding functions that have the same
3366 semantics as the standard library methods of the same names. These methods
3367 expose the underlying instruction list of a basic block in a way that is easy to
3368 manipulate. To get the full complement of container operations (including
3369 operations to update the list), you must use the <tt>getInstList()</tt>
3372 <li><tt>BasicBlock::InstListType &getInstList()</tt>
3374 <p>This method is used to get access to the underlying container that actually
3375 holds the Instructions. This method must be used when there isn't a forwarding
3376 function in the <tt>BasicBlock</tt> class for the operation that you would like
3377 to perform. Because there are no forwarding functions for "updating"
3378 operations, you need to use this if you want to update the contents of a
3379 <tt>BasicBlock</tt>.</p></li>
3381 <li><tt><a href="#Function">Function</a> *getParent()</tt>
3383 <p> Returns a pointer to <a href="#Function"><tt>Function</tt></a> the block is
3384 embedded into, or a null pointer if it is homeless.</p></li>
3386 <li><tt><a href="#TerminatorInst">TerminatorInst</a> *getTerminator()</tt>
3388 <p> Returns a pointer to the terminator instruction that appears at the end of
3389 the <tt>BasicBlock</tt>. If there is no terminator instruction, or if the last
3390 instruction in the block is not a terminator, then a null pointer is
3398 <!-- ======================================================================= -->
3399 <div class="doc_subsection">
3400 <a name="Argument">The <tt>Argument</tt> class</a>
3403 <div class="doc_text">
3405 <p>This subclass of Value defines the interface for incoming formal
3406 arguments to a function. A Function maintains a list of its formal
3407 arguments. An argument has a pointer to the parent Function.</p>
3411 <!-- *********************************************************************** -->
3414 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
3415 src="http://jigsaw.w3.org/css-validator/images/vcss-blue" alt="Valid CSS"></a>
3416 <a href="http://validator.w3.org/check/referer"><img
3417 src="http://www.w3.org/Icons/valid-html401" alt="Valid HTML 4.01 Strict"></a>
3419 <a href="mailto:dhurjati@cs.uiuc.edu">Dinakar Dhurjati</a> and
3420 <a href="mailto:sabre@nondot.org">Chris Lattner</a><br>
3421 <a href="http://llvm.org">The LLVM Compiler Infrastructure</a><br>
3422 Last modified: $Date$