| blob | 266c033bb4c4bc5674e00dd1d917145c5c823f98 |
3 <html>
4 <head>
8 </head>
9 <body>
11 <h1>
12 LLVM Programmer's Manual
13 </h1>
15 <ol>
18 <ul>
20 <!--
21 <li>The <tt>-time-passes</tt> option</li>
22 <li>How to use the LLVM Makefile system</li>
23 <li>How to write a regression test</li>
25 -->
26 </ul>
27 </li>
29 <ul>
34 <ul>
37 </ul>
38 </li>
40 option</a>
41 <ul>
44 </ul>
45 </li>
47 option</a></li>
48 <!--
49 <li>The <tt>InstVisitor</tt> template
50 <li>The general graph API
51 -->
53 </ul>
54 </li>
56 <ul>
58 <ul>
70 </ul></li>
72 <ul>
77 </ul></li>
79 <ul>
89 </ul></li>
91 <ul>
101 </ul></li>
103 <ul>
107 </ul></li>
108 </ul>
109 </li>
111 <ul>
113 <ul>
121 class pointer</a> </li>
123 complex example</a> </li>
125 the same way</a> </li>
127 use-def chains</a> </li>
129 successors of blocks</a></li>
130 </ul>
131 </li>
133 <ul>
140 </ul>
141 </li>
143 <!--
144 <li>Working with the Control Flow Graph
145 <ul>
146 <li>Accessing predecessors and successors of a <tt>BasicBlock</tt>
147 <li>
148 <li>
149 </ul>
150 -->
151 </ul>
152 </li>
155 <ul>
157 </a></li>
162 </ul>
163 </li>
166 <ul>
169 <li><a href="#UserLayout">The <tt>User</tt> and owned <tt>Use</tt> classes' memory layout</a></li>
170 </ul></li>
173 <ul>
177 <ul>
179 <ul>
182 <ul>
184 <ul>
187 </ul>
188 </li>
189 </ul>
190 </li>
191 </ul>
192 </li>
195 </ul>
196 </li>
197 </ul>
198 </li>
199 </ol>
208 </div>
210 <!-- *********************************************************************** -->
211 <h2>
213 </h2>
214 <!-- *********************************************************************** -->
216 <div>
218 <p>This document is meant to highlight some of the important classes and
219 interfaces available in the LLVM source-base. This manual is not
220 intended to explain what LLVM is, how it works, and what LLVM code looks
221 like. It assumes that you know the basics of LLVM and are interested
222 in writing transformations or otherwise analyzing or manipulating the
223 code.</p>
225 <p>This document should get you oriented so that you can find your
226 way in the continuously growing source code that makes up the LLVM
227 infrastructure. Note that this manual is not intended to serve as a
228 replacement for reading the source code, so if you think there should be
229 a method in one of these classes to do something, but it's not listed,
231 are provided to make this as easy as possible.</p>
233 <p>The first section of this document describes general information that is
234 useful to know when working in the LLVM infrastructure, and the second describes
235 the Core LLVM classes. In the future this manual will be extended with
236 information describing how to use extension libraries, such as dominator
237 information, CFG traversal routines, and useful utilities like the <tt><a
240 </div>
242 <!-- *********************************************************************** -->
243 <h2>
245 </h2>
246 <!-- *********************************************************************** -->
248 <div>
250 <p>This section contains general information that is useful if you are working
251 in the LLVM source-base, but that isn't specific to any particular API.</p>
253 <!-- ======================================================================= -->
254 <h3>
256 </h3>
258 <div>
260 <p>LLVM makes heavy use of the C++ Standard Template Library (STL),
261 perhaps much more than you are used to, or have seen before. Because of
262 this, you might want to do a little background reading in the
263 techniques used and capabilities of the library. There are many good
264 pages that discuss the STL, and several books on the subject that you
265 can get, so it will not be discussed in this document.</p>
269 <ol>
272 C++ Library reference</a> - an excellent reference for the STL and other parts
273 of the standard C++ library.</li>
276 O'Reilly book in the making. It has a decent Standard Library
277 Reference that rivals Dinkumware's, and is unfortunately no longer free since the
278 book has been published.</li>
281 Questions</a></li>
284 Contains a useful <a
289 Page</a></li>
293 the book).</a></li>
295 </ol>
299 to write maintainable code more than where to put your curly braces.</p>
301 </div>
303 <!-- ======================================================================= -->
304 <h3>
306 </h3>
308 <div>
310 <ol>
312 static and shared libraries across platforms</a></li>
313 </ol>
315 </div>
317 </div>
319 <!-- *********************************************************************** -->
320 <h2>
322 </h2>
323 <!-- *********************************************************************** -->
325 <div>
327 <p>Here we highlight some LLVM APIs that are generally useful and good to
328 know about when writing transformations.</p>
330 <!-- ======================================================================= -->
331 <h3>
334 </h3>
336 <div>
338 <p>The LLVM source-base makes extensive use of a custom form of RTTI.
340 operator, but they don't have some drawbacks (primarily stemming from
342 have a v-table). Because they are used so often, you must know what they
343 do and how they work. All of these templates are defined in the <a
345 file (note that you very rarely have to include this file directly).</p>
347 <dl>
351 "<tt>instanceof</tt>" operator. It returns true or false depending on whether
352 a reference or pointer points to an instance of the specified class. This can
353 be very useful for constraint checking of various sorts (example below).</p>
354 </dd>
359 converts a pointer or reference from a base class to a derived class, causing
360 an assertion failure if it is not really an instance of the right type. This
361 should be used in cases where you have some information that makes you believe
366 <pre>
368 if (isa<<a href="#Constant">Constant</a>>(V) || isa<<a href="#Argument">Argument</a>>(V) || isa<<a href="#GlobalValue">GlobalValue</a>>(V))
369 return true;
373 }
374 </pre>
375 </div>
379 operator.</p>
381 </dd>
386 It checks to see if the operand is of the specified type, and if so, returns a
387 pointer to it (this operator does not work with references). If the operand is
388 not of the correct type, a null pointer is returned. Thus, this works very
392 statement like this:</p>
395 <pre>
396 if (<a href="#AllocationInst">AllocationInst</a> *AI = dyn_cast<<a href="#AllocationInst">AllocationInst</a>>(Val)) {
398 }
399 </pre>
400 </div>
404 statement, which is very convenient.</p>
409 blocks to check for lots of different variants of classes. If you find
410 yourself wanting to do this, it is much cleaner and more efficient to use the
413 </dd>
419 argument (which it then propagates). This can sometimes be useful, allowing
420 you to combine several null checks into one.</p></dd>
426 as an argument (which it then propagates). This can sometimes be useful,
427 allowing you to combine several null checks into one.</p></dd>
429 </dl>
431 <p>These five templates can be used with any classes, whether they have a
432 v-table or not. To add support for these templates, you simply need to add
434 to. Describing this is currently outside the scope of this document, but there
435 are lots of examples in the LLVM source base.</p>
437 </div>
440 <!-- ======================================================================= -->
441 <h3>
444 </h3>
446 <div>
448 <p>Although LLVM generally does not do much string manipulation, we do have
449 several important APIs which take strings. Two important examples are the
450 Value class -- which has names for instructions, functions, etc. -- and the
451 StringMap class which is used extensively in LLVM and Clang.</p>
453 <p>These are generic classes, and they need to be able to accept strings which
454 may have embedded null characters. Therefore, they cannot simply take
456 clients to perform a heap allocation which is usually unnecessary. Instead,
458 passing strings efficiently.</p>
460 <!-- _______________________________________________________________________ -->
461 <h4>
463 </h4>
465 <div>
468 (a character array and a length) and supports the common operations available
471 <p>It can be implicitly constructed using a C style null-terminated string,
476 iterator find(StringRef Key);
477 </pre>
485 </pre>
491 for more information.</p>
494 pointers to external memory it is not generally safe to store an instance of the
495 class (unless you know that the external storage will not be freed). StringRef is
496 small and pervasive enough in LLVM that it should always be passed by value.</p>
498 </div>
500 <!-- _______________________________________________________________________ -->
501 <h4>
503 </h4>
505 <div>
508 strings. For example, a common LLVM paradigm is to name one instruction based on
509 the name of another instruction with a suffix, for example:</p>
512 <pre>
514 </pre>
515 </div>
519 which points to temporary (stack allocated) objects. Twines can be implicitly
520 constructed as the result of the plus operator applied to strings (i.e., a C
522 actual concatenation of strings until it is actually required, at which point
523 it can be efficiently rendered directly into a character array. This avoids
524 unnecessary heap allocation involved in constructing the temporary results of
525 string concatenation. See
527 for more information.</p>
530 and should almost never be stored or mentioned directly. They are intended
531 solely for use when defining a function which should be able to efficiently
532 accept concatenated strings.</p>
534 </div>
536 </div>
538 <!-- ======================================================================= -->
539 <h3>
541 </h3>
543 <div>
545 <p>Often when working on your pass you will put a bunch of debugging printouts
546 and other code into your pass. After you get it working, you want to remove
547 it, but you may need it again in the future (to work out new bugs that you run
548 across).</p>
550 <p> Naturally, because of this, you don't want to delete the debug printouts,
551 but you don't want them to always be noisy. A standard compromise is to comment
552 them out, allowing you to enable them if you need them in the future.</p>
556 this problem. Basically, you can put arbitrary code into the argument of the
561 <pre>
563 </pre>
564 </div>
569 <pre>
573 I am here!
574 </pre>
575 </div>
578 to not have to create "yet another" command line option for the debug output for
580 so they do not cause a performance impact at all (for the same reason, they
581 should also not contain side-effects!).</p>
584 enable or disable it directly in gdb. Just use "<tt>set DebugFlag=0</tt>" or
585 "<tt>set DebugFlag=1</tt>" from the gdb if the program is running. If the
586 program hasn't been started yet, you can always just run it with
589 <!-- _______________________________________________________________________ -->
590 <h4>
593 </h4>
595 <div>
599 generator). If you want to enable debug information with more fine-grained
601 option as follows:</p>
604 <pre>
605 #undef DEBUG_TYPE
607 #define DEBUG_TYPE "foo"
609 #undef DEBUG_TYPE
610 #define DEBUG_TYPE "bar"
612 #undef DEBUG_TYPE
613 #define DEBUG_TYPE ""
615 </pre>
616 </div>
621 <pre>
625 No debug type
626 'foo' debug type
627 'bar' debug type
628 No debug type (2)
630 'foo' debug type
632 'bar' debug type
633 </pre>
634 </div>
637 a file, to specify the debug type for the entire module (if you do this before
640 "bar", because there is no system in place to ensure that names do not
641 conflict. If two different modules use the same string, they will all be turned
642 on when the name is specified. This allows, for example, all debug information
644 even if the source lives in multiple files.</p>
648 statement. It takes an additional first parameter, which is the type to use. For
649 example, the preceding example could be written as:</p>
653 <pre>
658 </pre>
659 </div>
661 </div>
663 </div>
665 <!-- ======================================================================= -->
666 <h3>
668 option</a>
669 </h3>
671 <div>
676 keep track of what the LLVM compiler is doing and how effective various
677 optimizations are. It is useful to see what optimizations are contributing to
678 making a particular program run faster.</p>
680 <p>Often you may run your pass on some big program, and you're interested to see
681 how many times it makes a certain transformation. Although you can do this with
682 hand inspection, or some ad-hoc method, this is a real pain and not very useful
684 keep track of this information, and the calculated information is presented in a
685 uniform manner with the rest of the passes being executed.</p>
688 it are as follows:</p>
690 <ol>
694 <pre>
695 #define <a href="#DEBUG_TYPE">DEBUG_TYPE</a> "mypassname" <i>// This goes before any #includes.</i>
696 STATISTIC(NumXForms, "The # of times I did stuff");
697 </pre>
698 </div>
701 specified by the first argument. The pass name is taken from the DEBUG_TYPE
702 macro, and the description is taken from the second argument. The variable
708 <pre>
710 </pre>
711 </div>
713 </li>
714 </ol>
720 <pre>
723 </pre>
724 </div>
727 suite, it gives a report that looks like this:</p>
730 <pre>
731 7646 bitcodewriter - Number of normal instructions
732 725 bitcodewriter - Number of oversized instructions
733 129996 bitcodewriter - Number of bitcode bytes written
734 2817 raise - Number of insts DCEd or constprop'd
735 3213 raise - Number of cast-of-self removed
736 5046 raise - Number of expression trees converted
737 75 raise - Number of other getelementptr's formed
738 138 raise - Number of load/store peepholes
739 42 deadtypeelim - Number of unused typenames removed from symtab
740 392 funcresolve - Number of varargs functions resolved
741 27 globaldce - Number of global variables removed
742 2 adce - Number of basic blocks removed
743 134 cee - Number of branches revectored
744 49 cee - Number of setcc instruction eliminated
745 532 gcse - Number of loads removed
746 2919 gcse - Number of instructions removed
747 86 indvars - Number of canonical indvars added
748 87 indvars - Number of aux indvars removed
749 25 instcombine - Number of dead inst eliminate
750 434 instcombine - Number of insts combined
751 248 licm - Number of load insts hoisted
752 1298 licm - Number of insts hoisted to a loop pre-header
753 3 licm - Number of insts hoisted to multiple loop preds (bad, no loop pre-header)
754 75 mem2reg - Number of alloca's promoted
755 1444 cfgsimplify - Number of blocks simplified
756 </pre>
757 </div>
759 <p>Obviously, with so many optimizations, having a unified framework for this
760 stuff is very nice. Making your pass fit well into the framework makes it more
761 maintainable and useful.</p>
763 </div>
765 <!-- ======================================================================= -->
766 <h3>
768 </h3>
770 <div>
772 <p>Several of the important data structures in LLVM are graphs: for example
776 DAGs</a>. In many cases, while debugging various parts of the compiler, it is
777 nice to instantly visualize these graphs.</p>
779 <p>LLVM provides several callbacks that are available in a debug build to do
781 the current LLVM tool will pop up a window containing the CFG for the function
782 where each basic block is a node in the graph, and each node contains the
783 instructions in the block. Similarly, there also exists
788 up a window. Alternatively, you can sprinkle calls to these functions in your
789 code in places you want to debug.</p>
791 <p>Getting this to work requires a small amount of configuration. On Unix
793 toolkit, and make sure 'dot' and 'gv' are in your path. If you are running on
794 Mac OS/X, download and install the Mac OS/X <a
797 it) to your path. Once in your system and path are set up, rerun the LLVM
798 configure script and rebuild LLVM to enable this functionality.</p>
804 specified color (choices of colors can be found at <a
806 complex node attributes can be provided with <tt>call
809 Attributes</a>.) If you want to restart and clear all the current graph
812 <p>Note that graph visualization features are compiled out of Release builds
813 to reduce file size. This means that you need a Debug+Asserts or
814 Release+Asserts build to use these features.</p>
816 </div>
818 </div>
820 <!-- *********************************************************************** -->
821 <h2>
823 </h2>
824 <!-- *********************************************************************** -->
826 <div>
829 and we commonly use STL data structures. This section describes the trade-offs
830 you should consider when you pick one.</p>
832 <p>
833 The first step is a choose your own adventure: do you want a sequential
834 container, a set-like container, or a map-like container? The most important
835 thing when choosing a container is the algorithmic properties of how you plan to
836 access the container. Based on that, you should use:</p>
838 <ul>
840 of an value based on another value. Map-like containers also support
841 efficient queries for containment (whether a key is in the map). Map-like
842 containers generally do not support efficient reverse mapping (values to
843 keys). If you need that, use two maps. Some map-like containers also
844 support efficient iteration through the keys in sorted order. Map-like
845 containers are the most expensive sort, only use them if you need one of
846 these capabilities.</li>
849 stuff into a container that automatically eliminates duplicates. Some
850 set-like containers support efficient iteration through the elements in
851 sorted order. Set-like containers are more expensive than sequential
852 containers.
853 </li>
856 the most efficient way to add elements and keeps track of the order they are
857 added to the collection. They permit duplicates and support efficient
858 iteration, but do not support efficient look-up based on a key.
859 </li>
862 container or reference structure that is used for character or byte
863 arrays.</li>
866 perform set operations on sets of numeric id's, while automatically
867 eliminating duplicates. Bit containers require a maximum of 1 bit for each
868 identifier you want to store.
869 </li>
870 </ul>
872 <p>
873 Once the proper category of container is determined, you can fine tune the
874 memory use, constant factors, and cache behaviors of access by intelligently
875 picking a member of the category. Note that constant factors and cache behavior
876 can be a big deal. If you have a vector that usually only contains a few
877 elements (but could contain many), for example, it's much better to use
879 . Doing so avoids (relatively) expensive malloc/free calls, which dwarf the
880 cost of adding the elements to the container. </p>
882 </div>
885 <!-- ======================================================================= -->
886 <h3>
888 </h3>
890 <div>
891 There are a variety of sequential containers available for you, based on your
892 needs. Pick the first in this section that will do what you want.
894 <!-- _______________________________________________________________________ -->
895 <h4>
897 </h4>
899 <div>
900 <p>The llvm::ArrayRef class is the preferred class to use in an interface that
901 accepts a sequential list of elements in memory and just reads from them. By
902 taking an ArrayRef, the API can be passed a fixed size array, an std::vector,
903 an llvm::SmallVector and anything else that is contiguous in memory.
904 </p>
905 </div>
909 <!-- _______________________________________________________________________ -->
910 <h4>
912 </h4>
914 <div>
915 <p>Fixed size arrays are very simple and very fast. They are good if you know
916 exactly how many elements you have, or you have a (low) upper bound on how many
917 you have.</p>
918 </div>
920 <!-- _______________________________________________________________________ -->
921 <h4>
923 </h4>
925 <div>
926 <p>Heap allocated arrays (new[] + delete[]) are also simple. They are good if
927 the number of elements is variable, if you know how many elements you will need
928 before the array is allocated, and if the array is usually large (if not,
930 allocated array is the cost of the new/delete (aka malloc/free). Also note that
931 if you are allocating an array of a type with a constructor, the constructor and
932 destructors will be run for every element in the array (re-sizable vectors only
933 construct those elements actually used).</p>
934 </div>
936 <!-- _______________________________________________________________________ -->
937 <h4>
939 </h4>
942 <div>
944 that is optimized to avoid allocation in the case when a vector has zero or one
945 elements. It has two major restrictions: 1) it can only hold values of pointer
950 </div>
952 <!-- _______________________________________________________________________ -->
953 <h4>
955 </h4>
957 <div>
960 it supports efficient iteration, lays out elements in memory order (so you can
961 do pointer arithmetic between elements), supports efficient push_back/pop_back
962 operations, supports efficient random access to its elements, etc.</p>
964 <p>The advantage of SmallVector is that it allocates space for
966 the SmallVector is dynamically smaller than N, no malloc is performed. This can
967 be a big win in cases where the malloc/free call is far more expensive than the
968 code that fiddles around with the elements.</p>
971 predecessors/successors of a block is usually less than 8). On the other hand,
972 this makes the size of the SmallVector itself large, so you don't want to
973 allocate lots of them (doing so will waste a lot of space). As such,
974 SmallVectors are most useful when on the stack.</p>
976 <p>SmallVector also provides a nice portable and efficient replacement for
979 </div>
981 <!-- _______________________________________________________________________ -->
982 <h4>
984 </h4>
986 <div>
987 <p>
988 std::vector is well loved and respected. It is useful when SmallVector isn't:
989 when the size of the vector is often large (thus the small optimization will
990 rarely be a benefit) or if you will be allocating many instances of the vector
991 itself (which would waste space for elements that aren't in the container).
992 vector is also useful when interfacing with code that expects vectors :).
993 </p>
998 <pre>
999 for ( ... ) {
1001 use V;
1002 }
1003 </pre>
1004 </div>
1009 <pre>
1011 for ( ... ) {
1012 use V;
1013 V.clear();
1014 }
1015 </pre>
1016 </div>
1018 <p>Doing so will save (at least) one heap allocation and free per iteration of
1019 the loop.</p>
1021 </div>
1023 <!-- _______________________________________________________________________ -->
1024 <h4>
1026 </h4>
1028 <div>
1029 <p>std::deque is, in some senses, a generalized version of std::vector. Like
1030 std::vector, it provides constant time random access and other similar
1031 properties, but it also provides efficient access to the front of the list. It
1032 does not guarantee continuity of elements within memory.</p>
1034 <p>In exchange for this extra flexibility, std::deque has significantly higher
1035 constant factor costs than std::vector. If possible, use std::vector or
1036 something cheaper.</p>
1037 </div>
1039 <!-- _______________________________________________________________________ -->
1040 <h4>
1042 </h4>
1044 <div>
1045 <p>std::list is an extremely inefficient class that is rarely useful.
1046 It performs a heap allocation for every element inserted into it, thus having an
1047 extremely high constant factor, particularly for small data types. std::list
1048 also only supports bidirectional iteration, not random access iteration.</p>
1050 <p>In exchange for this high cost, std::list supports efficient access to both
1051 ends of the list (like std::deque, but unlike std::vector or SmallVector). In
1052 addition, the iterator invalidation characteristics of std::list are stronger
1053 than that of a vector class: inserting or removing an element into the list does
1054 not invalidate iterator or pointers to other elements in the list.</p>
1055 </div>
1057 <!-- _______________________________________________________________________ -->
1058 <h4>
1060 </h4>
1062 <div>
1064 intrusive, because it requires the element to store and provide access to the
1065 prev/next pointers for the list.</p>
1069 provides some novel characteristics. In particular, it can efficiently store
1070 polymorphic objects, the traits class is informed when an element is inserted or
1072 constant-time splice operation.</p>
1074 <p>These properties are exactly what we want for things like
1078 Related classes of interest are explained in the following subsections:
1079 <ul>
1084 </ul>
1085 </div>
1087 <!-- _______________________________________________________________________ -->
1088 <h4>
1090 </h4>
1092 <div>
1093 <p>
1094 Useful for storing a vector of values using only a few number of bits for each
1095 value. Apart from the standard operations of a vector-like container, it can
1096 also perform an 'or' set operation.
1097 </p>
1102 <pre>
1103 enum State {
1104 None = 0x0,
1105 FirstCondition = 0x1,
1106 SecondCondition = 0x2,
1107 Both = 0x3
1108 };
1110 State get() {
1112 Vec1.push_back(FirstCondition);
1115 Vec2.push_back(SecondCondition);
1117 Vec1 |= Vec2;
1118 return Vec1[0]; // returns 'Both'.
1119 }
1120 </pre>
1121 </div>
1123 </div>
1125 <!-- _______________________________________________________________________ -->
1126 <h4>
1128 </h4>
1130 <div>
1133 publicly derive from this traits class.</p>
1134 </div>
1136 <!-- _______________________________________________________________________ -->
1137 <h4>
1139 </h4>
1141 <div>
1143 supports a slightly narrower interface. Notably, inserters from
1147 used for a wide variety of customizations.</p>
1148 </div>
1150 <!-- _______________________________________________________________________ -->
1151 <h4>
1153 </h4>
1155 <div>
1158 in the default manner.</p>
1163 </div>
1165 <!-- _______________________________________________________________________ -->
1166 <h4>
1168 </h4>
1170 <div>
1172 citizen in the C++ ecosystem, it needs to support the standard container
1177 <p>The only sensible solution to this problem is to allocate a so-called
1179 iterator, providing the back-link to the last element. However conforming to the
1181 also must not be dereferenced.</p>
1184 how to allocate and store the sentinel. The corresponding policy is dictated
1186 whenever the need for a sentinel arises.</p>
1188 <p>While the default policy is sufficient in most cases, it may break down when
1191 is wasted. To alleviate the situation with numerous and voluminous
1197 arithmetic is used to obtain the sentinel, which is relative to the
1199 extra pointer, which serves as the back-link of the sentinel. This is the only
1200 field in the ghostly sentinel which can be legally accessed.</p>
1201 </div>
1203 <!-- _______________________________________________________________________ -->
1204 <h4>
1206 </h4>
1208 <div>
1211 <p>There are also various STL adapter classes such as std::queue,
1212 std::priority_queue, std::stack, etc. These provide simplified access to an
1213 underlying container but don't affect the cost of the container itself.</p>
1215 </div>
1216 </div>
1218 <!-- ======================================================================= -->
1219 <h3>
1221 </h3>
1223 <div>
1225 <p>
1226 There are a variety of ways to pass around and use strings in C and C++, and
1227 LLVM adds a few new options to choose from. Pick the first option on this list
1228 that will do what you need, they are ordered according to their relative cost.
1229 </p>
1230 <p>
1232 "<tt>const char*</tt>"'s. These have a number of problems, including the fact
1233 that they cannot represent embedded nul ("\0") characters, and do not have a
1234 length available efficiently. The general replacement for '<tt>const
1235 char*</tt>' is StringRef.
1236 </p>
1238 <p>For more information on choosing string containers for APIs, please see
1242 <!-- _______________________________________________________________________ -->
1243 <h4>
1245 </h4>
1247 <div>
1248 <p>
1249 The StringRef class is a simple value class that contains a pointer to a
1250 character and a length, and is quite related to the <a
1252 characters). Because StringRef carries a length with it, it safely handles
1253 strings with embedded nul characters in it, getting the length does not require
1254 a strlen call, and it even has very convenient APIs for slicing and dicing the
1255 character range that it represents.
1256 </p>
1258 <p>
1259 StringRef is ideal for passing simple strings around that are known to be live,
1260 either because they are C string literals, std::string, a C array, or a
1261 SmallVector. Each of these cases has an efficient implicit conversion to
1262 StringRef, which doesn't result in a dynamic strlen being executed.
1263 </p>
1265 <p>StringRef has a few major limitations which make more powerful string
1266 containers useful:</p>
1268 <ol>
1269 <li>You cannot directly convert a StringRef to a 'const char*' because there is
1270 no way to add a trailing nul (unlike the .c_str() method on various stronger
1271 classes).</li>
1274 <li>StringRef doesn't own or keep alive the underlying string bytes.
1275 As such it can easily lead to dangling pointers, and is not suitable for
1276 embedding in datastructures in most cases (instead, use an std::string or
1277 something like that).</li>
1279 <li>For the same reason, StringRef cannot be used as the return value of a
1280 method if the method "computes" the result string. Instead, use
1281 std::string.</li>
1283 <li>StringRef's do not allow you to mutate the pointed-to string bytes and it
1284 doesn't allow you to insert or remove bytes from the range. For editing
1285 operations like this, it interoperates with the <a
1287 </ol>
1289 <p>Because of its strengths and limitations, it is very common for a function to
1290 take a StringRef and for a method on an object to return a StringRef that
1291 points into some string that it owns.</p>
1293 </div>
1295 <!-- _______________________________________________________________________ -->
1296 <h4>
1298 </h4>
1300 <div>
1301 <p>
1302 The Twine class is used as an intermediary datatype for APIs that want to take
1303 a string that can be constructed inline with a series of concatenations.
1304 Twine works by forming recursive instances of the Twine datatype (a simple
1305 value object) on the stack as temporary objects, linking them together into a
1306 tree which is then linearized when the Twine is consumed. Twine is only safe
1307 to use as the argument to a function, and should always be a const reference,
1308 e.g.:
1309 </p>
1311 <pre>
1312 void foo(const Twine &T);
1313 ...
1314 StringRef X = ...
1315 unsigned i = ...
1316 foo(X + "." + Twine(i));
1317 </pre>
1320 together, and does not form intermediate strings containing "blarg" or
1321 "blarg.".
1322 </p>
1324 <p>Because Twine is constructed with temporary objects on the stack, and
1325 because these instances are destroyed at the end of the current statement,
1326 it is an inherently dangerous API. For example, this simple variant contains
1327 undefined behavior and will probably crash:</p>
1329 <pre>
1330 void foo(const Twine &T);
1331 ...
1332 StringRef X = ...
1333 unsigned i = ...
1335 foo(Tmp);
1336 </pre>
1338 <p>... because the temporaries are destroyed before the call. That said,
1339 Twine's are much more efficient than intermediate std::string temporaries, and
1340 they work really well with StringRef. Just be aware of their limitations.</p>
1342 </div>
1345 <!-- _______________________________________________________________________ -->
1346 <h4>
1348 </h4>
1350 <div>
1353 adds some convenience APIs like += that takes StringRef's. SmallString avoids
1354 allocating memory in the case when the preallocated space is enough to hold its
1355 data, and it calls back to general heap allocation when required. Since it owns
1356 its data, it is very safe to use and supports full mutation of the string.</p>
1358 <p>Like SmallVector's, the big downside to SmallString is their sizeof. While
1359 they are optimized for small strings, they themselves are not particularly
1360 small. This means that they work great for temporary scratch buffers on the
1361 stack, but should not generally be put into the heap: it is very rare to
1362 see a SmallString as the member of a frequently-allocated heap data structure
1363 or returned by-value.
1364 </p>
1366 </div>
1368 <!-- _______________________________________________________________________ -->
1369 <h4>
1371 </h4>
1373 <div>
1375 <p>The standard C++ std::string class is a very general class that (like
1376 SmallString) owns its underlying data. sizeof(std::string) is very reasonable
1377 so it can be embedded into heap data structures and returned by-value.
1378 On the other hand, std::string is highly inefficient for inline editing (e.g.
1379 concatenating a bunch of stuff together) and because it is provided by the
1380 standard library, its performance characteristics depend a lot of the host
1381 standard library (e.g. libc++ and MSVC provide a highly optimized string
1382 class, GCC contains a really slow implementation).
1383 </p>
1385 <p>The major disadvantage of std::string is that almost every operation that
1386 makes them larger can allocate memory, which is slow. As such, it is better
1387 to use SmallVector or Twine as a scratch buffer, but then use std::string to
1388 persist the result.</p>
1391 </div>
1393 <!-- end of strings -->
1394 </div>
1397 <!-- ======================================================================= -->
1398 <h3>
1400 </h3>
1402 <div>
1404 <p>Set-like containers are useful when you need to canonicalize multiple values
1405 into a single representation. There are several different choices for how to do
1406 this, providing various trade-offs.</p>
1408 <!-- _______________________________________________________________________ -->
1409 <h4>
1411 </h4>
1413 <div>
1415 <p>If you intend to insert a lot of elements, then do a lot of queries, a
1416 great approach is to use a vector (or other sequential container) with
1417 std::sort+std::unique to remove duplicates. This approach works really well if
1418 your usage pattern has these two distinct phases (insert then query), and can be
1420 </p>
1422 <p>
1423 This combination provides the several nice properties: the result data is
1424 contiguous in memory (good for cache locality), has few allocations, is easy to
1425 address (iterators in the final vector are just indices or pointers), and can be
1426 efficiently queried with a standard binary or radix search.</p>
1428 </div>
1430 <!-- _______________________________________________________________________ -->
1431 <h4>
1433 </h4>
1435 <div>
1437 <p>If you have a set-like data structure that is usually small and whose elements
1439 has space for N elements in place (thus, if the set is dynamically smaller than
1440 N, no malloc traffic is required) and accesses them with a simple linear search.
1441 When the set grows beyond 'N' elements, it allocates a more expensive representation that
1442 guarantees efficient access (for most types, it falls back to std::set, but for
1443 pointers it uses something far better, <a
1446 <p>The magic of this class is that it handles small sets extremely efficiently,
1447 but gracefully handles extremely large sets without loss of efficiency. The
1448 drawback is that the interface is quite small: it supports insertion, queries
1449 and erasing, but does not support iteration.</p>
1451 </div>
1453 <!-- _______________________________________________________________________ -->
1454 <h4>
1456 </h4>
1458 <div>
1460 <p>SmallPtrSet has all the advantages of <tt>SmallSet</tt> (and a <tt>SmallSet</tt> of pointers is
1462 more than 'N' insertions are performed, a single quadratically
1463 probed hash table is allocated and grows as needed, providing extremely
1464 efficient access (constant time insertion/deleting/queries with low constant
1465 factors) and is very stingy with malloc traffic.</p>
1468 whenever an insertion occurs. Also, the values visited by the iterators are not
1469 visited in sorted order.</p>
1471 </div>
1473 <!-- _______________________________________________________________________ -->
1474 <h4>
1476 </h4>
1478 <div>
1480 <p>
1481 DenseSet is a simple quadratically probed hash table. It excels at supporting
1482 small values: it uses a single allocation to hold all of the pairs that
1483 are currently inserted in the set. DenseSet is a great way to unique small
1484 values that are not simple pointers (use <a
1486 the same requirements for the value type that <a
1488 </p>
1490 </div>
1492 <!-- _______________________________________________________________________ -->
1493 <h4>
1495 </h4>
1497 <div>
1499 <p>
1500 FoldingSet is an aggregate class that is really good at uniquing
1501 expensive-to-create or polymorphic objects. It is a combination of a chained
1502 hash table with intrusive links (uniqued objects are required to inherit from
1504 its ID process.</p>
1507 a complex object (for example, a node in the code generator). The client has a
1508 description of *what* it wants to generate (it knows the opcode and all the
1509 operands), but we don't want to 'new' a node, then try inserting it into a set
1510 only to find out it already exists, at which point we would have to delete it
1511 and return the node that already exists.
1512 </p>
1514 <p>To support this style of client, FoldingSet perform a query with a
1515 FoldingSetNodeID (which wraps SmallVector) that can be used to describe the
1516 element that we want to query for. The query either returns the element
1517 matching the ID or it returns an opaque ID that indicates where insertion should
1518 take place. Construction of the ID usually does not require heap traffic.</p>
1520 <p>Because FoldingSet uses intrusive links, it can support polymorphic objects
1521 in the set (for example, you can have SDNode instances mixed with LoadSDNodes).
1522 Because the elements are individually allocated, pointers to the elements are
1523 stable: inserting or removing elements does not invalidate any pointers to other
1524 elements.
1525 </p>
1527 </div>
1529 <!-- _______________________________________________________________________ -->
1530 <h4>
1532 </h4>
1534 <div>
1537 many things but great at nothing. std::set allocates memory for each element
1538 inserted (thus it is very malloc intensive) and typically stores three pointers
1539 per element in the set (thus adding a large amount of per-element space
1540 overhead). It offers guaranteed log(n) performance, which is not particularly
1541 fast from a complexity standpoint (particularly if the elements of the set are
1542 expensive to compare, like strings), and has extremely high constant factors for
1543 lookup, insertion and removal.</p>
1545 <p>The advantages of std::set are that its iterators are stable (deleting or
1546 inserting an element from the set does not affect iterators or pointers to other
1547 elements) and that iteration over the set is guaranteed to be in sorted order.
1548 If the elements in the set are large, then the relative overhead of the pointers
1549 and malloc traffic is not a big deal, but if the elements of the set are small,
1550 std::set is almost never a good choice.</p>
1552 </div>
1554 <!-- _______________________________________________________________________ -->
1555 <h4>
1557 </h4>
1559 <div>
1562 Container</a>. The important property
1563 that this provides is efficient insertion with uniquing (duplicate elements are
1564 ignored) with iteration support. It implements this by inserting elements into
1565 both a set-like container and the sequential container, using the set-like
1566 container for uniquing and the sequential container for iteration.
1567 </p>
1569 <p>The difference between SetVector and other sets is that the order of
1570 iteration is guaranteed to match the order of insertion into the SetVector.
1571 This property is really important for things like sets of pointers. Because
1572 pointer values are non-deterministic (e.g. vary across runs of the program on
1573 different machines), iterating over the pointers in the set will
1574 not be in a well-defined order.</p>
1576 <p>
1577 The drawback of SetVector is that it requires twice as much space as a normal
1578 set and has the sum of constant factors from the set-like container and the
1579 sequential container that it uses. Use it *only* if you need to iterate over
1580 the elements in a deterministic order. SetVector is also expensive to delete
1581 elements out of (linear time), unless you use it's "pop_back" method, which is
1582 faster.
1583 </p>
1585 <p>SetVector is an adapter class that defaults to using std::vector and std::set
1586 for the underlying containers, so it is quite expensive. However,
1588 defaults to using a SmallVector and SmallSet of a specified size. If you use
1589 this, and if your sets are dynamically smaller than N, you will save a lot of
1590 heap traffic.</p>
1592 </div>
1594 <!-- _______________________________________________________________________ -->
1595 <h4>
1597 </h4>
1599 <div>
1601 <p>
1603 retains a unique ID for each element inserted into the set. It internally
1604 contains a map and a vector, and it assigns a unique ID for each value inserted
1605 into the set.</p>
1607 <p>UniqueVector is very expensive: its cost is the sum of the cost of
1608 maintaining both the map and vector, it has high complexity, high constant
1609 factors, and produces a lot of malloc traffic. It should be avoided.</p>
1611 </div>
1614 <!-- _______________________________________________________________________ -->
1615 <h4>
1617 </h4>
1619 <div>
1621 <p>
1622 The STL provides several other options, such as std::multiset and the various
1623 "hash_set" like containers (whether from C++ TR1 or from the SGI library). We
1624 never use hash_set and unordered_set because they are generally very expensive
1625 (each insertion requires a malloc) and very non-portable.
1626 </p>
1628 <p>std::multiset is useful if you're not interested in elimination of
1629 duplicates, but has all the drawbacks of std::set. A sorted vector (where you
1630 don't delete duplicate entries) or some other approach is almost always
1631 better.</p>
1633 </div>
1635 </div>
1637 <!-- ======================================================================= -->
1638 <h3>
1640 </h3>
1642 <div>
1643 Map-like containers are useful when you want to associate data to a key. As
1644 usual, there are a lot of different ways to do this. :)
1646 <!-- _______________________________________________________________________ -->
1647 <h4>
1649 </h4>
1651 <div>
1653 <p>
1654 If your usage pattern follows a strict insert-then-query approach, you can
1656 for set-like containers</a>. The only difference is that your query function
1657 (which uses std::lower_bound to get efficient log(n) lookup) should only compare
1658 the key, not both the key and value. This yields the same advantages as sorted
1659 vectors for sets.
1660 </p>
1661 </div>
1663 <!-- _______________________________________________________________________ -->
1664 <h4>
1666 </h4>
1668 <div>
1670 <p>
1671 Strings are commonly used as keys in maps, and they are difficult to support
1672 efficiently: they are variable length, inefficient to hash and compare when
1673 long, expensive to copy, etc. StringMap is a specialized container designed to
1674 cope with these issues. It supports mapping an arbitrary range of bytes to an
1675 arbitrary other object.</p>
1677 <p>The StringMap implementation uses a quadratically-probed hash table, where
1678 the buckets store a pointer to the heap allocated entries (and some other
1679 stuff). The entries in the map must be heap allocated because the strings are
1680 variable length. The string data (key) and the element object (value) are
1681 stored in the same allocation with the string data immediately after the element
1682 object. This container guarantees the "<tt>(char*)(&Value+1)</tt>" points
1683 to the key string for a value.</p>
1685 <p>The StringMap is very fast for several reasons: quadratic probing is very
1686 cache efficient for lookups, the hash value of strings in buckets is not
1687 recomputed when looking up an element, StringMap rarely has to touch the
1688 memory for unrelated objects when looking up a value (even when hash collisions
1689 happen), hash table growth does not recompute the hash values for strings
1690 already in the table, and each pair in the map is store in a single allocation
1691 (the string data is stored in the same allocation as the Value of a pair).</p>
1693 <p>StringMap also provides query methods that take byte ranges, so it only ever
1694 copies a string if a value is inserted into the table.</p>
1695 </div>
1697 <!-- _______________________________________________________________________ -->
1698 <h4>
1700 </h4>
1702 <div>
1703 <p>
1704 IndexedMap is a specialized container for mapping small dense integers (or
1705 values that can be mapped to small dense integers) to some other type. It is
1706 internally implemented as a vector with a mapping function that maps the keys to
1707 the dense integer range.
1708 </p>
1710 <p>
1711 This is useful for cases like virtual registers in the LLVM code generator: they
1712 have a dense mapping that is offset by a compile-time constant (the first
1713 virtual register ID).</p>
1715 </div>
1717 <!-- _______________________________________________________________________ -->
1718 <h4>
1720 </h4>
1722 <div>
1724 <p>
1725 DenseMap is a simple quadratically probed hash table. It excels at supporting
1726 small keys and values: it uses a single allocation to hold all of the pairs that
1727 are currently inserted in the map. DenseMap is a great way to map pointers to
1728 pointers, or map other small types to each other.
1729 </p>
1731 <p>
1732 There are several aspects of DenseMap that you should be aware of, however. The
1733 iterators in a densemap are invalidated whenever an insertion occurs, unlike
1734 map. Also, because DenseMap allocates space for a large number of key/value
1735 pairs (it starts with 64 by default), it will waste a lot of space if your keys
1736 or values are large. Finally, you must implement a partial specialization of
1737 DenseMapInfo for the key that you want, if it isn't already supported. This
1738 is required to tell DenseMap about two special marker values (which can never be
1739 inserted into the map) that it needs internally.</p>
1741 </div>
1743 <!-- _______________________________________________________________________ -->
1744 <h4>
1746 </h4>
1748 <div>
1750 <p>
1752 Value*s (or subclasses) to another type. When a Value is deleted or RAUW'ed,
1753 ValueMap will update itself so the new version of the key is mapped to the same
1754 value, just as if the key were a WeakVH. You can configure exactly how this
1755 happens, and what else happens on these two events, by passing
1758 </div>
1760 <!-- _______________________________________________________________________ -->
1761 <h4>
1763 </h4>
1765 <div>
1767 <p> IntervalMap is a compact map for small keys and values. It maps key
1768 intervals instead of single keys, and it will automatically coalesce adjacent
1769 intervals. When then map only contains a few intervals, they are stored in the
1770 map object itself to avoid allocations.</p>
1772 <p> The IntervalMap iterators are quite big, so they should not be passed around
1773 as STL iterators. The heavyweight iterators allow a smaller data structure.</p>
1775 </div>
1777 <!-- _______________________________________________________________________ -->
1778 <h4>
1780 </h4>
1782 <div>
1784 <p>
1786 a single allocation per pair inserted into the map, it offers log(n) lookup with
1787 an extremely large constant factor, imposes a space penalty of 3 pointers per
1788 pair in the map, etc.</p>
1790 <p>std::map is most useful when your keys or values are very large, if you need
1791 to iterate over the collection in sorted order, or if you need stable iterators
1792 into the map (i.e. they don't get invalidated if an insertion or deletion of
1793 another element takes place).</p>
1795 </div>
1797 <!-- _______________________________________________________________________ -->
1798 <h4>
1800 </h4>
1802 <div>
1804 <p>IntEqClasses provides a compact representation of equivalence classes of
1806 equivalence class. Classes can be joined by passing two class representatives to
1807 the join(a, b) method. Two integers are in the same class when findLeader()
1808 returns the same representative.</p>
1810 <p>Once all equivalence classes are formed, the map can be compressed so each
1812 is the total number of equivalence classes. The map must be uncompressed before
1813 it can be edited again.</p>
1815 </div>
1817 <!-- _______________________________________________________________________ -->
1818 <h4>
1820 </h4>
1822 <div>
1824 <p>
1825 The STL provides several other options, such as std::multimap and the various
1826 "hash_map" like containers (whether from C++ TR1 or from the SGI library). We
1827 never use hash_set and unordered_set because they are generally very expensive
1828 (each insertion requires a malloc) and very non-portable.</p>
1830 <p>std::multimap is useful if you want to map a key to multiple values, but has
1831 all the drawbacks of std::map. A sorted vector or some other approach is almost
1832 always better.</p>
1834 </div>
1836 </div>
1838 <!-- ======================================================================= -->
1839 <h3>
1841 </h3>
1843 <div>
1844 <p>Unlike the other containers, there are only two bit storage containers, and
1845 choosing when to use each is relatively straightforward.</p>
1847 <p>One additional option is
1849 implementation in many common compilers (e.g. commonly available versions of
1850 GCC) is extremely inefficient and 2) the C++ standards committee is likely to
1851 deprecate this container and/or change it significantly somehow. In any case,
1852 please don't use it.</p>
1854 <!-- _______________________________________________________________________ -->
1855 <h4>
1857 </h4>
1859 <div>
1860 <p> The BitVector container provides a dynamic size set of bits for manipulation.
1861 It supports individual bit setting/testing, as well as set operations. The set
1862 operations take time O(size of bitvector), but operations are performed one word
1863 at a time, instead of one bit at a time. This makes the BitVector very fast for
1864 set operations compared to other containers. Use the BitVector when you expect
1865 the number of set bits to be high (IE a dense set).
1866 </p>
1867 </div>
1869 <!-- _______________________________________________________________________ -->
1870 <h4>
1872 </h4>
1874 <div>
1875 <p> The SmallBitVector container provides the same interface as BitVector, but
1876 it is optimized for the case where only a small number of bits, less than
1877 25 or so, are needed. It also transparently supports larger bit counts, but
1878 slightly less efficiently than a plain BitVector, so SmallBitVector should
1879 only be used when larger counts are rare.
1880 </p>
1882 <p>
1883 At this time, SmallBitVector does not support set operations (and, or, xor),
1884 and its operator[] does not provide an assignable lvalue.
1885 </p>
1886 </div>
1888 <!-- _______________________________________________________________________ -->
1889 <h4>
1891 </h4>
1893 <div>
1894 <p> The SparseBitVector container is much like BitVector, with one major
1895 difference: Only the bits that are set, are stored. This makes the
1896 SparseBitVector much more space efficient than BitVector when the set is sparse,
1897 as well as making set operations O(number of set bits) instead of O(size of
1898 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
1899 (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).
1900 </p>
1901 </div>
1903 </div>
1905 </div>
1907 <!-- *********************************************************************** -->
1908 <h2>
1910 </h2>
1911 <!-- *********************************************************************** -->
1913 <div>
1915 <p>This section describes how to perform some very simple transformations of
1916 LLVM code. This is meant to give examples of common idioms used, showing the
1918 you should also read about the main classes that you will be working with. The
1920 and descriptions of the main classes that you should know about.</p>
1922 <!-- NOTE: this section should be heavy on example code -->
1923 <!-- ======================================================================= -->
1924 <h3>
1926 </h3>
1928 <div>
1930 <p>The LLVM compiler infrastructure have many different data structures that may
1931 be traversed. Following the example of the C++ standard template library, the
1932 techniques used to traverse these various data structures are all basically the
1935 function returns an iterator pointing to one past the last valid element of the
1937 between the two operations.</p>
1939 <p>Because the pattern for iteration is common across many different aspects of
1940 the program representation, the standard template library algorithms may be used
1941 on them, and it is easier to remember how to iterate. First we show a few common
1942 examples of the data structures that need to be traversed. Other data
1943 structures are traversed in very similar ways.</p>
1945 <!-- _______________________________________________________________________ -->
1946 <h4>
1950 </h4>
1952 <div>
1955 transform in some way; in particular, you'd like to manipulate its
1962 <pre>
1969 </pre>
1970 </div>
1972 <p>Note that i can be used as if it were a pointer for the purposes of
1974 because the indirection operator is overloaded for the iterator
1978 </div>
1980 <!-- _______________________________________________________________________ -->
1981 <h4>
1985 </h4>
1987 <div>
1990 easy to iterate over the individual instructions that make up
1995 <pre>
2001 </pre>
2002 </div>
2004 <p>However, this isn't really the best way to print out the contents of a
2006 anything you'll care about, you could have just invoked the print routine on the
2009 </div>
2011 <!-- _______________________________________________________________________ -->
2012 <h4>
2016 </h4>
2018 <div>
2025 small example that shows how to dump all instructions in a function to the standard error stream:<p>
2028 <pre>
2032 for (inst_iterator I = inst_begin(F), E = inst_end(F); I != E; ++I)
2034 </pre>
2035 </div>
2038 work list with its initial contents. For example, if you wanted to
2040 F, all you would need to do is something like:</p>
2043 <pre>
2047 for (inst_iterator I = inst_begin(F), E = inst_end(F); I != E; ++I)
2048 worklist.insert(&*I);
2049 </pre>
2050 </div>
2055 </div>
2057 <!-- _______________________________________________________________________ -->
2058 <h4>
2060 vice-versa)</a>
2061 </h4>
2063 <div>
2065 <p>Sometimes, it'll be useful to grab a reference (or pointer) to a class
2066 instance when all you've got at hand is an iterator. Well, extracting
2067 a reference or a pointer from an iterator is very straight-forward.
2072 <pre>
2075 const Instruction& inst = *j;
2076 </pre>
2077 </div>
2079 <p>However, the iterators you'll be working with in the LLVM framework are
2080 special: they will automatically convert to a ptr-to-instance type whenever they
2081 need to. Instead of dereferencing the iterator and then taking the address of
2082 the result, you can simply assign the iterator to the proper pointer type and
2083 you get the dereference and address-of operation as a result of the assignment
2084 (behind the scenes, this is a result of overloading casting mechanisms). Thus
2085 the last line of the last example,</p>
2088 <pre>
2089 Instruction *pinst = &*i;
2090 </pre>
2091 </div>
2096 <pre>
2097 Instruction *pinst = i;
2098 </pre>
2099 </div>
2101 <p>It's also possible to turn a class pointer into the corresponding iterator,
2102 and this is a constant time operation (very efficient). The following code
2103 snippet illustrates use of the conversion constructors provided by LLVM
2104 iterators. By using these, you can explicitly grab the iterator of something
2105 without actually obtaining it via iteration over some structure:</p>
2108 <pre>
2109 void printNextInstruction(Instruction* inst) {
2110 BasicBlock::iterator it(inst);
2113 }
2114 </pre>
2115 </div>
2117 <p>Unfortunately, these implicit conversions come at a cost; they prevent
2118 these iterators from conforming to standard iterator conventions, and thus
2119 from being usable with standard algorithms and containers. For example, they
2121 from compiling:</p>
2124 <pre>
2126 </pre>
2127 </div>
2129 <p>Because of this, these implicit conversions may be removed some day,
2132 </div>
2134 <!--_______________________________________________________________________-->
2135 <h4>
2137 example</a>
2138 </h4>
2140 <div>
2142 <p>Say that you're writing a FunctionPass and would like to count all the
2146 much more straight-forward manner, but this example will allow us to explore how
2148 is what we want to do:</p>
2151 <pre>
2152 initialize callCounter to zero
2153 for each Function f in the Module
2154 for each BasicBlock b in f
2155 for each Instruction i in b
2156 if (i is a CallInst and calls the given function)
2157 increment callCounter
2158 </pre>
2159 </div>
2161 <p>And the actual code is (remember, because we're writing a
2166 <pre>
2167 Function* targetFunc = ...;
2169 class OurFunctionPass : public FunctionPass {
2170 public:
2171 OurFunctionPass(): callCounter(0) { }
2173 virtual runOnFunction(Function& F) {
2174 for (Function::iterator b = F.begin(), be = F.end(); b != be; ++b) {
2181 if (callInst->getCalledFunction() == targetFunc)
2182 ++callCounter;
2183 }
2184 }
2185 }
2186 }
2188 private:
2189 unsigned callCounter;
2190 };
2191 </pre>
2192 </div>
2194 </div>
2196 <!--_______________________________________________________________________-->
2197 <h4>
2199 </h4>
2201 <div>
2203 <p>You may have noticed that the previous example was a bit oversimplified in
2204 that it did not deal with call sites generated by 'invoke' instructions. In
2205 this, and in other situations, you may find that you want to treat
2208 less closely-related things. For these cases, LLVM provides a handy wrapper
2209 class called <a
2216 reference and it should not be dynamically allocated or deallocated using
2218 assignable and constructable, with costs equivalents to that of a bare pointer.
2219 If you look at its definition, it has only a single pointer member.</p>
2221 </div>
2223 <!--_______________________________________________________________________-->
2224 <h4>
2226 </h4>
2228 <div>
2240 <pre>
2241 Function *F = ...;
2247 }
2248 </pre>
2249 </div>
2253 doing it only once in the loop body and reusing its result.</p>
2260 all of the values that a particular instruction uses (that is, the operands of
2264 <pre>
2265 Instruction *pi = ...;
2268 Value *v = *i;
2270 }
2271 </pre>
2272 </div>
2275 mutation free algorithms (such as analyses, etc.). For this purpose above
2282 </div>
2284 <!--_______________________________________________________________________-->
2285 <h4>
2287 successors of blocks</a>
2288 </h4>
2290 <div>
2292 <p>Iterating over the predecessors and successors of a block is quite easy
2294 this to iterate over all predecessors of BB:</p>
2297 <pre>
2298 #include "llvm/Support/CFG.h"
2299 BasicBlock *BB = ...;
2301 for (pred_iterator PI = pred_begin(BB), E = pred_end(BB); PI != E; ++PI) {
2302 BasicBlock *Pred = *PI;
2304 }
2305 </pre>
2306 </div>
2308 <p>Similarly, to iterate over successors use
2309 succ_iterator/succ_begin/succ_end.</p>
2311 </div>
2313 </div>
2315 <!-- ======================================================================= -->
2316 <h3>
2318 </h3>
2320 <div>
2322 <p>There are some primitive transformation operations present in the LLVM
2323 infrastructure that are worth knowing about. When performing
2324 transformations, it's fairly common to manipulate the contents of basic
2325 blocks. This section describes some of the common methods for doing so
2326 and gives example code.</p>
2328 <!--_______________________________________________________________________-->
2329 <h4>
2332 </h4>
2334 <div>
2339 constructor for the kind of instruction to instantiate and provide the necessary
2344 <pre>
2345 AllocaInst* ai = new AllocaInst(Type::Int32Ty);
2346 </pre>
2347 </div>
2351 subclass is likely to have varying default parameters which change the semantics
2352 of the instruction, so refer to the <a
2358 <p>It is very useful to name the values of instructions when you're able to, as
2359 this facilitates the debugging of your transformations. If you end up looking
2360 at generated LLVM machine code, you definitely want to have logical names
2361 associated with the results of instructions! By supplying a value for the
2363 associate a logical name with the result of the instruction's execution at
2364 run time. For example, say that I'm writing a transformation that dynamically
2365 allocates space for an integer on the stack, and that integer is going to be
2366 used as some kind of index by some other code. To accomplish this, I place an
2372 <pre>
2374 </pre>
2375 </div>
2378 execution value, which is a pointer to an integer on the run time stack.</p>
2385 <ul>
2386 <li>Insertion into an explicit instruction list
2393 <pre>
2394 BasicBlock *pb = ...;
2395 Instruction *pi = ...;
2396 Instruction *newInst = new Instruction(...);
2399 </pre>
2400 </div>
2404 classes provide constructors which take a pointer to a
2406 looked like: </p>
2409 <pre>
2410 BasicBlock *pb = ...;
2411 Instruction *newInst = new Instruction(...);
2414 </pre>
2415 </div>
2420 <pre>
2421 BasicBlock *pb = ...;
2422 Instruction *newInst = new Instruction(..., pb);
2423 </pre>
2424 </div>
2426 <p>which is much cleaner, especially if you are creating
2427 long instruction streams.</p></li>
2429 <li>Insertion into an implicit instruction list
2432 are implicitly associated with an existing instruction list: the instruction
2433 list of the enclosing basic block. Thus, we could have accomplished the same
2435 </p>
2438 <pre>
2439 Instruction *pi = ...;
2440 Instruction *newInst = new Instruction(...);
2443 </pre>
2444 </div>
2446 <p>In fact, this sequence of steps occurs so frequently that the
2448 constructors which take (as a default parameter) a pointer to an
2452 provided instruction, immediately before that instruction. Using an
2454 parameter, the above code becomes:</p>
2457 <pre>
2458 Instruction* pi = ...;
2459 Instruction* newInst = new Instruction(..., pi);
2460 </pre>
2461 </div>
2463 <p>which is much cleaner, especially if you're creating a lot of
2465 </ul>
2467 </div>
2469 <!--_______________________________________________________________________-->
2470 <h4>
2472 </h4>
2474 <div>
2476 <p>Deleting an instruction from an existing sequence of instructions that form a
2478 call the instruction's eraseFromParent() method. For example:</p>
2481 <pre>
2483 I->eraseFromParent();
2484 </pre>
2485 </div>
2487 <p>This unlinks the instruction from its containing basic block and deletes
2488 it. If you'd just like to unlink the instruction from its containing basic
2491 </div>
2493 <!--_______________________________________________________________________-->
2494 <h4>
2497 </h4>
2499 <div>
2503 <p>Including "<a href="/doxygen/BasicBlockUtils_8h-source.html">llvm/Transforms/Utils/BasicBlockUtils.h</a>"
2509 <ul>
2512 <p>This function replaces all uses of a given instruction with a value,
2513 and then removes the original instruction. The following example
2514 illustrates the replacement of the result of a particular
2516 pointer to an integer.</p>
2519 <pre>
2520 AllocaInst* instToReplace = ...;
2521 BasicBlock::iterator ii(instToReplace);
2524 Constant::getNullValue(PointerType::getUnqual(Type::Int32Ty)));
2525 </pre></div></li>
2529 <p>This function replaces a particular instruction with another
2530 instruction, inserting the new instruction into the basic block at the
2531 location where the old instruction was, and replacing any uses of the old
2532 instruction with the new instruction. The following example illustrates
2536 <pre>
2537 AllocaInst* instToReplace = ...;
2538 BasicBlock::iterator ii(instToReplace);
2542 </pre></div></li>
2543 </ul>
2551 information.</p>
2553 <!-- Value::replaceAllUsesWith User::replaceUsesOfWith Point out:
2554 include/llvm/Transforms/Utils/ especially BasicBlockUtils.h with:
2555 ReplaceInstWithValue, ReplaceInstWithInst -->
2557 </div>
2559 <!--_______________________________________________________________________-->
2560 <h4>
2562 </h4>
2564 <div>
2566 <p>Deleting a global variable from a module is just as easy as deleting an
2567 Instruction. First, you must have a pointer to the global variable that you wish
2568 to delete. You use this pointer to erase it from its parent, the module.
2569 For example:</p>
2572 <pre>
2575 GV->eraseFromParent();
2576 </pre>
2577 </div>
2579 </div>
2581 </div>
2583 <!-- ======================================================================= -->
2584 <h3>
2586 </h3>
2588 <div>
2590 <p>In generating IR, you may need some complex types. If you know these types
2593 has two forms depending on whether you're building types for cross-compilation
2596 out of types from
2598 namespace and pointers, functions, arrays, etc. built of
2600 whose size may depend on the host compiler. For example,</p>
2603 <pre>
2605 </pre>
2606 </div>
2611 <pre>
2613 params.push_back(PointerType::getUnqual(Type::Int32Ty));
2614 FunctionType *ft = FunctionType::get(Type::Int8Ty, params, false);
2615 </pre>
2616 </div>
2621 </div>
2623 </div>
2625 <!-- *********************************************************************** -->
2626 <h2>
2628 </h2>
2629 <!-- *********************************************************************** -->
2631 <div>
2632 <p>
2633 This section describes the interaction of the LLVM APIs with multithreading,
2634 both on the part of client applications, and in the JIT, in the hosted
2635 application.
2636 </p>
2638 <p>
2639 Note that LLVM's support for multithreading is still relatively young. Up
2640 through version 2.5, the execution of threaded hosted applications was
2641 supported, but not threaded client access to the APIs. While this use case is
2643 ensure proper operation in multithreaded mode.
2644 </p>
2646 <p>
2647 Note that, on Unix-like platforms, LLVM requires the presence of GCC's atomic
2648 intrinsics in order to support threaded operation. If you need a
2649 multhreading-capable LLVM on a platform without a suitably modern system
2650 compiler, consider compiling LLVM and LLVM-GCC in single-threaded mode, and
2651 using the resultant compiler to build a copy of LLVM with multithreading
2652 support.
2653 </p>
2655 <!-- ======================================================================= -->
2656 <h3>
2658 </h3>
2660 <div>
2662 <p>
2663 In order to properly protect its internal data structures while avoiding
2664 excessive locking overhead in the single-threaded case, the LLVM must intialize
2665 certain data structures necessary to provide guards around its internals. To do
2667 making any concurrent LLVM API calls. To subsequently tear down these
2670 mode.
2671 </p>
2673 <p>
2675 say that no other LLVM API calls may be executing at any time during the
2677 </tt>. It's is the client's responsibility to enforce this isolation.
2678 </p>
2680 <p>
2682 failure of the initialization. Failure typically indicates that your copy of
2683 LLVM was built without multithreading support, typically because GCC atomic
2684 intrinsics were not found in your system compiler. In this case, the LLVM API
2687 must be taken</a> to ensure that side exits and the like do not accidentally
2688 result in concurrent LLVM API calls.
2689 </p>
2690 </div>
2692 <!-- ======================================================================= -->
2693 <h3>
2695 </h3>
2697 <div>
2698 <p>
2700 to deallocate memory used for internal structures. This will also invoke
2704 </p>
2706 <p>
2707 Note that, if you use scope-based shutdown, you can use the
2709 destructor.
2710 </div>
2712 <!-- ======================================================================= -->
2713 <h3>
2715 </h3>
2717 <div>
2718 <p>
2720 initialization of static resources, such as the global type tables. Before the
2723 however, it uses double-checked locking to implement thread-safe lazy
2724 initialization.
2725 </p>
2727 <p>
2728 Note that, because no other threads are allowed to issue LLVM API calls before
2731 </p>
2733 <p>
2735 APIs provide access to the global lock used to implement the double-checked
2736 locking for lazy initialization. These should only be used internally to LLVM,
2737 and only if you know what you're doing!
2738 </p>
2739 </div>
2741 <!-- ======================================================================= -->
2742 <h3>
2744 </h3>
2746 <div>
2747 <p>
2749 to operate multiple, isolated instances of LLVM concurrently within the same
2750 address space. For instance, in a hypothetical compile-server, the compilation
2751 of an individual translation unit is conceptually independent from all the
2752 others, and it would be desirable to be able to compile incoming translation
2753 units concurrently on independent server threads. Fortunately,
2755 </p>
2757 <p>
2764 safe to compile on multiple threads simultaneously, as long as no two threads
2765 operate on entities within the same context.
2766 </p>
2768 <p>
2769 In practice, very few places in the API require the explicit specification of a
2772 other entities can determine what context they belong to by looking at their
2774 maintain this interface design.
2775 </p>
2777 <p>
2781 isolation is not a concern.
2782 </p>
2783 </div>
2785 <!-- ======================================================================= -->
2786 <h3>
2788 </h3>
2790 <div>
2791 <p>
2792 LLVM's "eager" JIT compiler is safe to use in threaded programs. Multiple
2795 run code output by the JIT concurrently. The user must still ensure that only
2797 might be modifying it. One way to do that is to always hold the JIT lock while
2801 </p>
2803 <p>When the JIT is configured to compile lazily (using
2806 updating call sites after a function is lazily-jitted. It's still possible to
2807 use the lazy JIT in a threaded program if you ensure that only one thread at a
2808 time can call any particular lazy stub and that the JIT lock guards any IR
2809 access, but we suggest using only the eager JIT in threaded programs.
2810 </p>
2811 </div>
2813 </div>
2815 <!-- *********************************************************************** -->
2816 <h2>
2818 </h2>
2819 <!-- *********************************************************************** -->
2821 <div>
2822 <p>
2823 This section describes some of the advanced or obscure API's that most clients
2824 do not need to be aware of. These API's tend manage the inner workings of the
2825 LLVM system, and only need to be accessed in unusual circumstances.
2826 </p>
2829 <!-- ======================================================================= -->
2830 <h3>
2832 </h3>
2834 <div>
2840 </p>
2843 by most clients. It should only be used when iteration over the symbol table
2844 names themselves are required, which is very special purpose. Note that not
2845 all LLVM
2847 an empty name) do not exist in the symbol table.
2848 </p>
2850 <p>Symbol tables support iteration over the values in the symbol
2855 appropriate symbol table.</p>
2857 </div>
2861 <!-- ======================================================================= -->
2862 <h3>
2864 </h3>
2866 <div>
2872 addition and removal.</p>
2874 <!-- ______________________________________________________________________ -->
2875 <h4>
2878 </a>
2879 </h4>
2881 <div>
2882 <p>
2884 or refer to them out-of-line by means of a pointer. A mixed variant
2887 </p>
2889 <p>
2891 <ul>
2892 <li><p>Layout a)
2894 object and there are a fixed number of them.</p>
2896 <li><p>Layout b)
2899 number of them.</p>
2900 </ul>
2901 <p>
2902 As of v2.4 each layout still possesses a direct pointer to the
2904 we stick to this redundancy for the sake of simplicity.
2906 has. (Theoretically this information can also be calculated
2907 given the scheme presented below.)</p>
2908 <p>
2910 enforce the following memory layouts:</p>
2912 <ul>
2913 <li><p>Layout a) is modelled by prepending the <tt>User</tt> object by the <tt>Use[]</tt> array.</p>
2915 <pre>
2916 ...---.---.---.---.-------...
2917 | P | P | P | P | User
2918 '''---'---'---'---'-------'''
2919 </pre>
2922 <pre>
2923 .-------...
2924 | User
2925 '-------'''
2926 |
2927 v
2928 .---.---.---.---...
2929 | P | P | P | P |
2930 '---'---'---'---'''
2931 </pre>
2932 </ul>
2936 </div>
2938 <!-- ______________________________________________________________________ -->
2939 <h4>
2941 </h4>
2943 <div>
2944 <p>
2947 recover it. This is accomplished by the following scheme:</p>
2949 A bit-encoding in the 2 LSBits (least significant bits) of the <tt>Use::Prev</tt> allows to find the
2951 <ul>
2956 </ul>
2957 <p>
2960 we have to walk to the next stop picking up digits
2961 and calculating the offset:</p>
2962 <pre>
2963 .---.---.---.---.---.---.---.---.---.---.---.---.---.---.---.---.----------------
2965 '---'---'---'---'---'---'---'---'---'---'---'---'---'---'---'---'----------------
2967 | | | | |__>
2968 | | | |__________>
2969 | | |______________________>
2970 | |______________________________________>
2971 |__________________________________________________________>
2972 </pre>
2973 <p>
2974 Only the significant number of bits need to be stored between the
2978 </div>
2980 <!-- ______________________________________________________________________ -->
2981 <h4>
2983 </h4>
2985 <div>
2986 <p>
2987 The following literate Haskell fragment demonstrates the concept:</p>
2990 <pre>
2991 > import Test.QuickCheck
2992 >
2997 >
3003 >
3004 > takeLast n ss = reverse $ take n $ reverse ss
3005 >
3007 >
3008 </pre>
3009 </div>
3010 <p>
3012 <p>
3013 The reverse algorithm computes the length of the string just by examining
3014 a certain prefix:</p>
3017 <pre>
3022 >
3025 > decode walk acc _ = walk + acc
3026 >
3027 </pre>
3028 </div>
3029 <p>
3031 <p>
3035 <pre>
3037 > testcaseLength = length testcase
3038 >
3040 > where arr = takeLast n testcase
3041 >
3042 </pre>
3043 </div>
3044 <p>
3047 <pre>
3048 *Main> quickCheck identityProp
3049 OK, passed 100 tests.
3050 </pre>
3051 <p>
3052 Let's be a bit more exhaustive:</p>
3055 <pre>
3056 >
3058 >
3059 </pre>
3060 </div>
3061 <p>
3064 <pre>
3065 *Main> deepCheck identityProp
3066 OK, passed 500 tests.
3067 </pre>
3069 </div>
3071 <!-- ______________________________________________________________________ -->
3072 <h4>
3074 </h4>
3076 <div>
3078 <p>
3082 tag bits.</p>
3083 <p>
3087 the LSBit set. (Portability is relying on the fact that all known compilers place the
3090 </div>
3092 </div>
3094 </div>
3096 <!-- *********************************************************************** -->
3097 <h2>
3099 </h2>
3100 <!-- *********************************************************************** -->
3102 <div>
3106 <p>The Core LLVM classes are the primary means of representing the program
3107 being inspected or transformed. The core LLVM classes are defined in
3111 <!-- ======================================================================= -->
3112 <h3>
3114 </h3>
3116 <div>
3122 subclasses. They are hidden because they offer no useful functionality beyond
3126 named, but this is not a requirement. There exists exactly
3127 one instance of a given shape at any one time. This allows type equality to
3128 be performed with address equality of the Type Instance. That is, given two
3130 </p>
3132 <!-- _______________________________________________________________________ -->
3133 <h4>
3135 </h4>
3137 <div>
3139 <ul>
3143 floating point types.</li>
3146 that don't have a size are abstract types, labels and void.</li>
3148 </ul>
3149 </div>
3151 <!-- _______________________________________________________________________ -->
3152 <h4>
3154 </h4>
3155 <div>
3156 <dl>
3158 <dd>Subclass of DerivedType that represents integer types of any bit width.
3161 <ul>
3163 type of a specific bit width.</li>
3165 type.</li>
3166 </ul>
3167 </dd>
3169 <dd>This is subclassed by ArrayType, PointerType and VectorType.
3170 <ul>
3172 of the elements in the sequential type. </li>
3173 </ul>
3174 </dd>
3176 <dd>This is a subclass of SequentialType and defines the interface for array
3177 types.
3178 <ul>
3180 elements in the array. </li>
3181 </ul>
3182 </dd>
3186 <dd>Subclass of SequentialType for vector types. A
3187 vector type is similar to an ArrayType but is distinguished because it is
3188 a first class type whereas ArrayType is not. Vector types are used for
3189 vector operations and are usually small vectors of of an integer or floating
3190 point type.</dd>
3194 <dd>Subclass of DerivedTypes for function types.
3195 <ul>
3197 function</li>
3199 return type of the function.</li>
3201 the type of the ith parameter.</li>
3203 number of formal parameters.</li>
3204 </ul>
3205 </dd>
3206 </dl>
3207 </div>
3209 </div>
3211 <!-- ======================================================================= -->
3212 <h3>
3214 </h3>
3216 <div>
3223 programs. An LLVM module is effectively either a translation unit of the
3224 original program or a combination of several translation units merged by the
3229 helpful member functions that try to make common operations easy.</p>
3231 <!-- _______________________________________________________________________ -->
3232 <h4>
3234 </h4>
3236 <div>
3238 <ul>
3240 </ul>
3243 provide a name for it (probably based on the name of the translation unit).</p>
3245 <ul>
3252 <p>These are forwarding methods that make it easy to access the contents of
3254 list.</p></li>
3259 necessary to use when you need to update the list or perform a complex
3260 action that doesn't have a forwarding method.</p>
3263 </ul>
3265 <hr>
3267 <ul>
3275 <p> These are forwarding methods that make it easy to access the contents of
3283 use when you need to update the list or perform a complex action that
3284 doesn't have a forwarding method.</p>
3287 </ul>
3289 <hr>
3291 <ul>
3298 </ul>
3300 <hr>
3302 <ul>
3315 external declaration for the function and return it.</p></li>
3322 string.</p></li>
3329 name, true is returned and the <a
3331 </ul>
3333 </div>
3335 </div>
3337 <!-- ======================================================================= -->
3338 <h3>
3340 </h3>
3342 <div>
3345 <br>
3349 base. It represents a typed value that may be used (among other things) as an
3357 for a program. For example, an incoming argument to a function (represented
3359 every instruction in the function that references the argument. To keep track
3365 methods, shown below.</p>
3369 method. In addition, all LLVM values can be named. The "name" of the
3373 <pre>
3375 </pre>
3376 </div>
3379 that the name of any value may be missing (an empty string), so names should
3381 debugging printouts), they should not be used to keep track of values or map
3385 <p>One important aspect of LLVM is that there is no distinction between an SSA
3386 variable and the operation that produces it. Because of this, any reference to
3387 the value produced by an instruction (or the value available as an incoming
3388 argument, for example) is represented as a direct pointer to the instance of
3389 the class that
3390 represents this value. Although this may take some getting used to, it
3391 simplifies the representation and makes it easier to manipulate.</p>
3393 <!-- _______________________________________________________________________ -->
3394 <h4>
3396 </h4>
3398 <div>
3400 <ul>
3402 use-list<br>
3404 the use-list<br>
3406 value.<br>
3409 the use-list.<br>
3411 use-list.<br>
3413 element in the list.
3414 <p> These methods are the interface to access the def-use
3415 information in LLVM. As with all other iterators in LLVM, the naming
3417 </li>
3420 </li>
3426 </li>
3431 "<tt>V</tt>" instead. For example, if you detect that an instruction always
3432 produces a constant value (for example through constant folding), you can
3433 replace all uses of the instruction with the constant like this:</p>
3436 <pre>
3437 Inst->replaceAllUsesWith(ConstVal);
3438 </pre>
3439 </div>
3441 </ul>
3443 </div>
3445 </div>
3447 <!-- ======================================================================= -->
3448 <h3>
3450 </h3>
3452 <div>
3454 <p>
3467 Single Assignment (SSA) form, there can only be one definition referred to,
3468 allowing this direct connection. This connection provides the use-def
3469 information in LLVM.</p>
3471 <!-- _______________________________________________________________________ -->
3472 <h4>
3474 </h4>
3476 <div>
3479 an index access interface and through an iterator based interface.</p>
3481 <ul>
3485 convenient form for direct access.</p></li>
3488 list<br>
3490 the operand list.<br>
3492 operand list.
3493 <p> Together, these methods make up the iterator based interface to
3495 </ul>
3497 </div>
3499 </div>
3501 <!-- ======================================================================= -->
3502 <h3>
3504 </h3>
3506 <div>
3515 instructions. It provides only a few methods, but is a very commonly used
3517 opcode (instruction type) and the parent <a
3519 into. To represent a specific type of instruction, one of many subclasses of
3528 file contains some meta-data about the various different types of instructions
3529 in LLVM. It describes the enum values that are used as opcodes (for example
3534 this file confuses doxygen, so these enum values don't show up correctly in the
3537 <!-- _______________________________________________________________________ -->
3538 <h4>
3541 </a>
3542 </h4>
3543 <div>
3544 <ul>
3546 <p>This subclasses represents all two operand instructions whose operands
3547 must be the same type, except for the comparison instructions.</p></li>
3550 common operations on cast instructions.</p>
3552 <p>This subclass respresents the two comparison instructions,
3556 <p>This subclass is the parent of all terminator instructions (those which
3557 can terminate a block).</p>
3558 </ul>
3559 </div>
3561 <!-- _______________________________________________________________________ -->
3562 <h4>
3565 </a>
3566 </h4>
3568 <div>
3570 <ul>
3575 <p>Returns true if the instruction writes to memory, i.e. it is a
3580 <p>Returns another instance of the specified instruction, identical
3581 in all ways to the original except that the instruction has no parent
3583 and it has no name</p></li>
3584 </ul>
3586 </div>
3588 </div>
3590 <!-- ======================================================================= -->
3591 <h3>
3593 </h3>
3595 <div>
3597 <p>Constant represents a base class for different types of constants. It
3598 is subclassed by ConstantInt, ConstantArray, etc. for representing
3600 a subclass, which represents the address of a global variable or function.
3601 </p>
3603 <!-- _______________________________________________________________________ -->
3605 <div>
3606 <ul>
3607 <li>ConstantInt : This subclass of Constant represents an integer constant of
3608 any width.
3609 <ul>
3611 value of this constant, an APInt value.</li>
3613 value to an int64_t via sign extension. If the value (not the bit width)
3614 of the APInt is too large to fit in an int64_t, an assertion will result.
3615 For this reason, use of this method is discouraged.</li>
3617 value to a uint64_t via zero extension. IF the value (not the bit width)
3618 of the APInt is too large to fit in a uint64_t, an assertion will result.
3619 For this reason, use of this method is discouraged.</li>
3622 The type is implied as the IntegerType that corresponds to the bit width
3625 Returns the ConstantInt object that represents the value provided by
3627 </ul>
3628 </li>
3629 <li>ConstantFP : This class represents a floating point constant.
3630 <ul>
3632 this constant. </li>
3633 </ul>
3634 </li>
3635 <li>ConstantArray : This represents a constant array.
3636 <ul>
3638 a vector of component constants that makeup this array. </li>
3639 </ul>
3640 </li>
3641 <li>ConstantStruct : This represents a constant struct.
3642 <ul>
3644 a vector of component constants that makeup this array. </li>
3645 </ul>
3646 </li>
3647 <li>GlobalValue : This represents either a global variable or a function. In
3648 either case, the value is a constant fixed address (after linking).
3649 </li>
3650 </ul>
3651 </div>
3653 </div>
3655 <!-- ======================================================================= -->
3656 <h3>
3658 </h3>
3660 <div>
3665 Class</a><br>
3672 Because they are visible at global scope, they are also subject to linking with
3673 other globals defined in different translation units. To control the linking
3680 unit, and does not participate in linking. If it has external linkage, it is
3681 visible to external code, and does participate in linking. In addition to
3687 global is always a pointer to its contents. It is important to remember this
3692 the address of the first element of this array and the value of the
3700 <!-- _______________________________________________________________________ -->
3701 <h4>
3704 </a>
3705 </h4>
3707 <div>
3709 <ul>
3714 <p> </p>
3715 </li>
3718 GlobalValue is currently embedded into.</p></li>
3719 </ul>
3721 </div>
3723 </div>
3725 <!-- ======================================================================= -->
3726 <h3>
3728 </h3>
3730 <div>
3741 actually one of the more complex classes in the LLVM hierarchy because it must
3749 ordering of the blocks in the function, which indicate how the code will be
3750 laid out by the backend. Additionally, the first <a
3753 block. There are no implicit exit nodes, and in fact there may be multiple exit
3757 function hasn't been linked in yet.</p>
3767 LLVM feature that is only used when you have to look up a value by name. Aside
3769 internally to make sure that there are not conflicts between the names of <a
3776 is its address (after linking) which is guaranteed to be constant.</p>
3778 <!-- _______________________________________________________________________ -->
3779 <h4>
3782 </a>
3783 </h4>
3785 <div>
3787 <ul>
3792 the the program. The constructor must specify the type of the function to
3793 create and what type of linkage the function should have. The <a
3795 specifies the formal arguments and return value for the function. The same
3798 in which the function is defined. If this argument is provided, the function
3799 will automatically be inserted into that module's list of
3800 functions.</p></li>
3805 function is "external", it does not have a body, and thus must be resolved
3806 by linking with a function defined in a different translation unit.</p></li>
3814 <p>These are forwarding methods that make it easy to access the contents of
3816 list.</p></li>
3821 is necessary to use when you need to update the list or perform a complex
3822 action that doesn't have a forwarding method.</p></li>
3825 iterator<br>
3831 <p>These are forwarding methods that make it easy to access the contents of
3833 list.</p></li>
3838 necessary to use when you need to update the list or perform a complex
3839 action that doesn't have a forwarding method.</p></li>
3844 function. Because the entry block for the function is always the first
3853 function.</p></li>
3859 </ul>
3861 </div>
3863 </div>
3865 <!-- ======================================================================= -->
3866 <h3>
3868 </h3>
3870 <div>
3874 <br>
3876 Class</a><br>
3882 <p>Global variables are represented with the (surprise surprise)
3885 always referenced by their address (global values must live in memory, so their
3886 "name" refers to their constant address). See
3888 variables may have an initial value (which must be a
3890 they may be marked as "constant" themselves (indicating that their contents
3891 never change at runtime).</p>
3893 <!-- _______________________________________________________________________ -->
3894 <h4>
3897 </a>
3898 </h4>
3900 <div>
3902 <ul>
3907 <p>Create a new global variable of the specified type. If
3909 unchanging for the program. The Linkage parameter specifies the type of
3910 linkage (internal, external, weak, linkonce, appending) for the variable.
3911 If the linkage is InternalLinkage, WeakAnyLinkage, WeakODRLinkage,
3912 LinkOnceAnyLinkage or LinkOnceODRLinkage, then the resultant
3913 global variable will have internal linkage. AppendingLinkage concatenates
3914 together all instances (in different translation units) of the variable
3915 into a single variable but is only applicable to arrays. See
3917 further details on linkage types. Optionally an initializer, a name, and the
3918 module to put the variable into may be specified for the global variable as
3919 well.</p></li>
3923 <p>Returns true if this is a global variable that is known not to
3924 be modified at runtime.</p></li>
3933 to call this method if there is no initializer.</p></li>
3934 </ul>
3936 </div>
3938 </div>
3940 <!-- ======================================================================= -->
3941 <h3>
3943 </h3>
3945 <div>
3950 Class</a><br>
3953 <p>This class represents a single entry single exit section of the code,
3954 commonly known as a basic block by the compiler community. The
3957 Matching the language definition, the last element of this list of instructions
3958 is always a terminator instruction (a subclass of the <a
3961 <p>In addition to tracking the list of instructions that make up the block, the
3970 <!-- _______________________________________________________________________ -->
3971 <h4>
3974 </a>
3975 </h4>
3977 <div>
3978 <ul>
3984 insertion into a function. The constructor optionally takes a name for the new
3987 automatically inserted at the end of the specified <a
3995 STL-style functions for accessing the instruction list.
3997 <p>These methods and typedefs are forwarding functions that have the same
3998 semantics as the standard library methods of the same names. These methods
3999 expose the underlying instruction list of a basic block in a way that is easy to
4000 manipulate. To get the full complement of container operations (including
4002 method.</p></li>
4006 <p>This method is used to get access to the underlying container that actually
4007 holds the Instructions. This method must be used when there isn't a forwarding
4009 to perform. Because there are no forwarding functions for "updating"
4010 operations, you need to use this if you want to update the contents of a
4016 embedded into, or a null pointer if it is homeless.</p></li>
4020 <p> Returns a pointer to the terminator instruction that appears at the end of
4022 instruction in the block is not a terminator, then a null pointer is
4023 returned.</p></li>
4025 </ul>
4027 </div>
4029 </div>
4031 <!-- ======================================================================= -->
4032 <h3>
4034 </h3>
4036 <div>
4038 <p>This subclass of Value defines the interface for incoming formal
4039 arguments to a function. A Function maintains a list of its formal
4040 arguments. An argument has a pointer to the parent Function.</p>
4042 </div>
4044 </div>
4046 <!-- *********************************************************************** -->
4047 <hr>
4048 <address>
4057 Last modified: $Date$
4058 </address>
4060 </body>
4061 </html>