1 <!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01//EN"
2 "http://www.w3.org/TR/html4/strict.dtd">
5 <meta http-equiv="Content-Type" content="text/html; charset=utf-8">
6 <title>Writing an LLVM Pass</title>
7 <link rel="stylesheet" href="llvm.css" type="text/css">
11 <div class="doc_title">
16 <li><a href="#introduction">Introduction - What is a pass?</a></li>
17 <li><a href="#quickstart">Quick Start - Writing hello world</a>
19 <li><a href="#makefile">Setting up the build environment</a></li>
20 <li><a href="#basiccode">Basic code required</a></li>
21 <li><a href="#running">Running a pass with <tt>opt</tt></a></li>
23 <li><a href="#passtype">Pass classes and requirements</a>
25 <li><a href="#ImmutablePass">The <tt>ImmutablePass</tt> class</a></li>
26 <li><a href="#ModulePass">The <tt>ModulePass</tt> class</a>
28 <li><a href="#runOnModule">The <tt>runOnModule</tt> method</a></li>
30 <li><a href="#CallGraphSCCPass">The <tt>CallGraphSCCPass</tt> class</a>
32 <li><a href="#doInitialization_scc">The <tt>doInitialization(CallGraph
33 &)</tt> method</a></li>
34 <li><a href="#runOnSCC">The <tt>runOnSCC</tt> method</a></li>
35 <li><a href="#doFinalization_scc">The <tt>doFinalization(CallGraph
36 &)</tt> method</a></li>
38 <li><a href="#FunctionPass">The <tt>FunctionPass</tt> class</a>
40 <li><a href="#doInitialization_mod">The <tt>doInitialization(Module
41 &)</tt> method</a></li>
42 <li><a href="#runOnFunction">The <tt>runOnFunction</tt> method</a></li>
43 <li><a href="#doFinalization_mod">The <tt>doFinalization(Module
44 &)</tt> method</a></li>
46 <li><a href="#LoopPass">The <tt>LoopPass</tt> class</a>
48 <li><a href="#doInitialization_loop">The <tt>doInitialization(Loop *,
49 LPPassManager &)</tt> method</a></li>
50 <li><a href="#runOnLoop">The <tt>runOnLoop</tt> method</a></li>
51 <li><a href="#doFinalization_loop">The <tt>doFinalization()
54 <li><a href="#BasicBlockPass">The <tt>BasicBlockPass</tt> class</a>
56 <li><a href="#doInitialization_fn">The <tt>doInitialization(Function
57 &)</tt> method</a></li>
58 <li><a href="#runOnBasicBlock">The <tt>runOnBasicBlock</tt>
60 <li><a href="#doFinalization_fn">The <tt>doFinalization(Function
61 &)</tt> method</a></li>
63 <li><a href="#MachineFunctionPass">The <tt>MachineFunctionPass</tt>
66 <li><a href="#runOnMachineFunction">The
67 <tt>runOnMachineFunction(MachineFunction &)</tt> method</a></li>
70 <li><a href="#registration">Pass Registration</a>
72 <li><a href="#print">The <tt>print</tt> method</a></li>
74 <li><a href="#interaction">Specifying interactions between passes</a>
76 <li><a href="#getAnalysisUsage">The <tt>getAnalysisUsage</tt>
78 <li><a href="#AU::addRequired">The <tt>AnalysisUsage::addRequired<></tt> and <tt>AnalysisUsage::addRequiredTransitive<></tt> methods</a></li>
79 <li><a href="#AU::addPreserved">The <tt>AnalysisUsage::addPreserved<></tt> method</a></li>
80 <li><a href="#AU::examples">Example implementations of <tt>getAnalysisUsage</tt></a></li>
81 <li><a href="#getAnalysis">The <tt>getAnalysis<></tt> and <tt>getAnalysisToUpdate<></tt> methods</a></li>
83 <li><a href="#analysisgroup">Implementing Analysis Groups</a>
85 <li><a href="#agconcepts">Analysis Group Concepts</a></li>
86 <li><a href="#registerag">Using <tt>RegisterAnalysisGroup</tt></a></li>
88 <li><a href="#passStatistics">Pass Statistics</a>
89 <li><a href="#passmanager">What PassManager does</a>
91 <li><a href="#releaseMemory">The <tt>releaseMemory</tt> method</a></li>
93 <li><a href="#registering">Registering dynamically loaded passes</a>
95 <li><a href="#registering_existing">Using existing registries</a></li>
96 <li><a href="#registering_new">Creating new registries</a></li>
98 <li><a href="#debughints">Using GDB with dynamically loaded passes</a>
100 <li><a href="#breakpoint">Setting a breakpoint in your pass</a></li>
101 <li><a href="#debugmisc">Miscellaneous Problems</a></li>
103 <li><a href="#future">Future extensions planned</a>
105 <li><a href="#SMP">Multithreaded LLVM</a></li>
106 <li><a href="#PassFunctionPass"><tt>ModulePass</tt>es requiring
107 <tt>FunctionPass</tt>es</a></li>
111 <div class="doc_author">
112 <p>Written by <a href="mailto:sabre@nondot.org">Chris Lattner</a> and
113 <a href="mailto:jlaskey@mac.com">Jim Laskey</a></p>
116 <!-- *********************************************************************** -->
117 <div class="doc_section">
118 <a name="introduction">Introduction - What is a pass?</a>
120 <!-- *********************************************************************** -->
122 <div class="doc_text">
124 <p>The LLVM Pass Framework is an important part of the LLVM system, because LLVM
125 passes are where most of the interesting parts of the compiler exist. Passes
126 perform the transformations and optimizations that make up the compiler, they
127 build the analysis results that are used by these transformations, and they are,
128 above all, a structuring technique for compiler code.</p>
130 <p>All LLVM passes are subclasses of the <tt><a
131 href="http://llvm.org/doxygen/classllvm_1_1Pass.html">Pass</a></tt>
132 class, which implement functionality by overriding virtual methods inherited
133 from <tt>Pass</tt>. Depending on how your pass works, you should inherit from
134 the <tt><a href="#ModulePass">ModulePass</a></tt>, <tt><a
135 href="#CallGraphSCCPass">CallGraphSCCPass</a></tt>, <tt><a
136 href="#FunctionPass">FunctionPass</a></tt>, or <tt><a
137 href="#LoopPass">LoopPass</a></tt>, or <tt><a
138 href="#BasicBlockPass">BasicBlockPass</a></tt> classes, which gives the system
139 more information about what your pass does, and how it can be combined with
140 other passes. One of the main features of the LLVM Pass Framework is that it
141 schedules passes to run in an efficient way based on the constraints that your
142 pass meets (which are indicated by which class they derive from).</p>
144 <p>We start by showing you how to construct a pass, everything from setting up
145 the code, to compiling, loading, and executing it. After the basics are down,
146 more advanced features are discussed.</p>
150 <!-- *********************************************************************** -->
151 <div class="doc_section">
152 <a name="quickstart">Quick Start - Writing hello world</a>
154 <!-- *********************************************************************** -->
156 <div class="doc_text">
158 <p>Here we describe how to write the "hello world" of passes. The "Hello" pass
159 is designed to simply print out the name of non-external functions that exist in
160 the program being compiled. It does not modify the program at all, it just
161 inspects it. The source code and files for this pass are available in the LLVM
162 source tree in the <tt>lib/Transforms/Hello</tt> directory.</p>
166 <!-- ======================================================================= -->
167 <div class="doc_subsection">
168 <a name="makefile">Setting up the build environment</a>
171 <div class="doc_text">
173 <p>First, you need to create a new directory somewhere in the LLVM source
174 base. For this example, we'll assume that you made
175 <tt>lib/Transforms/Hello</tt>. Next, you must set up a build script
176 (Makefile) that will compile the source code for the new pass. To do this,
177 copy the following into <tt>Makefile</tt>:</p>
180 <div class="doc_code"><pre>
181 # Makefile for hello pass
183 # Path to top level of LLVM heirarchy
186 # Name of the library to build
189 # Make the shared library become a loadable module so the tools can
190 # dlopen/dlsym on the resulting library.
193 # Tell the build system which LLVM libraries your pass needs. You'll probably
194 # need at least LLVMSystem.a, LLVMSupport.a, LLVMCore.a but possibly several
196 LLVMLIBS = LLVMCore.a LLVMSupport.a LLVMSystem.a
198 # Include the makefile implementation stuff
199 include $(LEVEL)/Makefile.common
202 <p>This makefile specifies that all of the <tt>.cpp</tt> files in the current
203 directory are to be compiled and linked together into a
204 <tt>Debug/lib/Hello.so</tt> shared object that can be dynamically loaded by
205 the <tt>opt</tt> or <tt>bugpoint</tt> tools via their <tt>-load</tt> options.
206 If your operating system uses a suffix other than .so (such as windows or
207 Mac OS/X), the appropriate extension will be used.</p>
209 <p>Now that we have the build scripts set up, we just need to write the code for
214 <!-- ======================================================================= -->
215 <div class="doc_subsection">
216 <a name="basiccode">Basic code required</a>
219 <div class="doc_text">
221 <p>Now that we have a way to compile our new pass, we just have to write it.
224 <div class="doc_code"><pre>
225 <b>#include</b> "<a href="http://llvm.org/doxygen/Pass_8h-source.html">llvm/Pass.h</a>"
226 <b>#include</b> "<a href="http://llvm.org/doxygen/Function_8h-source.html">llvm/Function.h</a>"
229 <p>Which are needed because we are writing a <tt><a
230 href="http://llvm.org/doxygen/classllvm_1_1Pass.html">Pass</a></tt>, and
231 we are operating on <tt><a
232 href="http://llvm.org/doxygen/classllvm_1_1Function.html">Function</a></tt>'s.</p>
235 <div class="doc_code"><pre>
236 <b>using namespace llvm;</b>
238 <p>... which is required because the functions from the include files
239 live in the llvm namespace.
244 <div class="doc_code"><pre>
248 <p>... which starts out an anonymous namespace. Anonymous namespaces are to C++
249 what the "<tt>static</tt>" keyword is to C (at global scope). It makes the
250 things declared inside of the anonymous namespace only visible to the current
251 file. If you're not familiar with them, consult a decent C++ book for more
254 <p>Next, we declare our pass itself:</p>
256 <div class="doc_code"><pre>
257 <b>struct</b> Hello : <b>public</b> <a href="#FunctionPass">FunctionPass</a> {
260 <p>This declares a "<tt>Hello</tt>" class that is a subclass of <tt><a
261 href="http://llvm.org/doxygen/classllvm_1_1FunctionPass.html">FunctionPass</a></tt>.
262 The different builtin pass subclasses are described in detail <a
263 href="#passtype">later</a>, but for now, know that <a
264 href="#FunctionPass"><tt>FunctionPass</tt></a>'s operate a function at a
267 <div class="doc_code"><pre>
268 <b>virtual bool</b> <a href="#runOnFunction">runOnFunction</a>(Function &F) {
269 llvm::cerr << "<i>Hello: </i>" << F.getName() << "\n";
272 }; <i>// end of struct Hello</i>
275 <p>We declare a "<a href="#runOnFunction"><tt>runOnFunction</tt></a>" method,
276 which overloads an abstract virtual method inherited from <a
277 href="#FunctionPass"><tt>FunctionPass</tt></a>. This is where we are supposed
278 to do our thing, so we just print out our message with the name of each
281 <div class="doc_code"><pre>
282 RegisterPass<Hello> X("<i>hello</i>", "<i>Hello World Pass</i>");
283 } <i>// end of anonymous namespace</i>
286 <p>Lastly, we <a href="#registration">register our class</a> <tt>Hello</tt>,
287 giving it a command line
288 argument "<tt>hello</tt>", and a name "<tt>Hello World Pass</tt>".</p>
290 <p>As a whole, the <tt>.cpp</tt> file looks like:</p>
292 <div class="doc_code"><pre>
293 <b>#include</b> "<a href="http://llvm.org/doxygen/Pass_8h-source.html">llvm/Pass.h</a>"
294 <b>#include</b> "<a href="http://llvm.org/doxygen/Function_8h-source.html">llvm/Function.h</a>"
296 <b>using namespace llvm;</b>
299 <b>struct Hello</b> : <b>public</b> <a href="#FunctionPass">FunctionPass</a> {
300 <b>virtual bool</b> <a href="#runOnFunction">runOnFunction</a>(Function &F) {
301 llvm::cerr << "<i>Hello: </i>" << F.getName() << "\n";
306 RegisterPass<Hello> X("<i>hello</i>", "<i>Hello World Pass</i>");
310 <p>Now that it's all together, compile the file with a simple "<tt>gmake</tt>"
311 command in the local directory and you should get a new
312 "<tt>Debug/lib/Hello.so</tt> file. Note that everything in this file is
313 contained in an anonymous namespace: this reflects the fact that passes are self
314 contained units that do not need external interfaces (although they can have
315 them) to be useful.</p>
319 <!-- ======================================================================= -->
320 <div class="doc_subsection">
321 <a name="running">Running a pass with <tt>opt</tt></a>
324 <div class="doc_text">
326 <p>Now that you have a brand new shiny shared object file, we can use the
327 <tt>opt</tt> command to run an LLVM program through your pass. Because you
328 registered your pass with the <tt>RegisterPass</tt> template, you will be able to
329 use the <tt>opt</tt> tool to access it, once loaded.</p>
331 <p>To test it, follow the example at the end of the <a
332 href="GettingStarted.html">Getting Started Guide</a> to compile "Hello World" to
333 LLVM. We can now run the bytecode file (<tt>hello.bc</tt>) for the program
334 through our transformation like this (or course, any bytecode file will
337 <div class="doc_code"><pre>
338 $ opt -load ../../../Debug/lib/Hello.so -hello < hello.bc > /dev/null
344 <p>The '<tt>-load</tt>' option specifies that '<tt>opt</tt>' should load your
345 pass as a shared object, which makes '<tt>-hello</tt>' a valid command line
346 argument (which is one reason you need to <a href="#registration">register your
347 pass</a>). Because the hello pass does not modify the program in any
348 interesting way, we just throw away the result of <tt>opt</tt> (sending it to
349 <tt>/dev/null</tt>).</p>
351 <p>To see what happened to the other string you registered, try running
352 <tt>opt</tt> with the <tt>--help</tt> option:</p>
354 <div class="doc_code"><pre>
355 $ opt -load ../../../Debug/lib/Hello.so --help
356 OVERVIEW: llvm .bc -> .bc modular optimizer
358 USAGE: opt [options] <input bytecode>
361 Optimizations available:
363 -funcresolve - Resolve Functions
364 -gcse - Global Common Subexpression Elimination
365 -globaldce - Dead Global Elimination
366 <b>-hello - Hello World Pass</b>
367 -indvars - Canonicalize Induction Variables
368 -inline - Function Integration/Inlining
369 -instcombine - Combine redundant instructions
373 <p>The pass name get added as the information string for your pass, giving some
374 documentation to users of <tt>opt</tt>. Now that you have a working pass, you
375 would go ahead and make it do the cool transformations you want. Once you get
376 it all working and tested, it may become useful to find out how fast your pass
377 is. The <a href="#passManager"><tt>PassManager</tt></a> provides a nice command
378 line option (<tt>--time-passes</tt>) that allows you to get information about
379 the execution time of your pass along with the other passes you queue up. For
382 <div class="doc_code"><pre>
383 $ opt -load ../../../Debug/lib/Hello.so -hello -time-passes < hello.bc > /dev/null
387 ===============================================================================
388 ... Pass execution timing report ...
389 ===============================================================================
390 Total Execution Time: 0.02 seconds (0.0479059 wall clock)
392 ---User Time--- --System Time-- --User+System-- ---Wall Time--- --- Pass Name ---
393 0.0100 (100.0%) 0.0000 ( 0.0%) 0.0100 ( 50.0%) 0.0402 ( 84.0%) Bytecode Writer
394 0.0000 ( 0.0%) 0.0100 (100.0%) 0.0100 ( 50.0%) 0.0031 ( 6.4%) Dominator Set Construction
395 0.0000 ( 0.0%) 0.0000 ( 0.0%) 0.0000 ( 0.0%) 0.0013 ( 2.7%) Module Verifier
396 <b> 0.0000 ( 0.0%) 0.0000 ( 0.0%) 0.0000 ( 0.0%) 0.0033 ( 6.9%) Hello World Pass</b>
397 0.0100 (100.0%) 0.0100 (100.0%) 0.0200 (100.0%) 0.0479 (100.0%) TOTAL
400 <p>As you can see, our implementation above is pretty fast :). The additional
401 passes listed are automatically inserted by the '<tt>opt</tt>' tool to verify
402 that the LLVM emitted by your pass is still valid and well formed LLVM, which
403 hasn't been broken somehow.</p>
405 <p>Now that you have seen the basics of the mechanics behind passes, we can talk
406 about some more details of how they work and how to use them.</p>
410 <!-- *********************************************************************** -->
411 <div class="doc_section">
412 <a name="passtype">Pass classes and requirements</a>
414 <!-- *********************************************************************** -->
416 <div class="doc_text">
418 <p>One of the first things that you should do when designing a new pass is to
419 decide what class you should subclass for your pass. The <a
420 href="#basiccode">Hello World</a> example uses the <tt><a
421 href="#FunctionPass">FunctionPass</a></tt> class for its implementation, but we
422 did not discuss why or when this should occur. Here we talk about the classes
423 available, from the most general to the most specific.</p>
425 <p>When choosing a superclass for your Pass, you should choose the <b>most
426 specific</b> class possible, while still being able to meet the requirements
427 listed. This gives the LLVM Pass Infrastructure information necessary to
428 optimize how passes are run, so that the resultant compiler isn't unneccesarily
433 <!-- ======================================================================= -->
434 <div class="doc_subsection">
435 <a name="ImmutablePass">The <tt>ImmutablePass</tt> class</a>
438 <div class="doc_text">
440 <p>The most plain and boring type of pass is the "<tt><a
441 href="http://llvm.org/doxygen/classllvm_1_1ImmutablePass.html">ImmutablePass</a></tt>"
442 class. This pass type is used for passes that do not have to be run, do not
443 change state, and never need to be updated. This is not a normal type of
444 transformation or analysis, but can provide information about the current
445 compiler configuration.</p>
447 <p>Although this pass class is very infrequently used, it is important for
448 providing information about the current target machine being compiled for, and
449 other static information that can affect the various transformations.</p>
451 <p><tt>ImmutablePass</tt>es never invalidate other transformations, are never
452 invalidated, and are never "run".</p>
456 <!-- ======================================================================= -->
457 <div class="doc_subsection">
458 <a name="ModulePass">The <tt>ModulePass</tt> class</a>
461 <div class="doc_text">
464 href="http://llvm.org/doxygen/classllvm_1_1ModulePass.html">ModulePass</a></tt>"
465 class is the most general of all superclasses that you can use. Deriving from
466 <tt>ModulePass</tt> indicates that your pass uses the entire program as a unit,
467 refering to function bodies in no predictable order, or adding and removing
468 functions. Because nothing is known about the behavior of <tt>ModulePass</tt>
469 subclasses, no optimization can be done for their execution. A module pass
470 can use function level passes (e.g. dominators) using getAnalysis interface
471 <tt> getAnalysis<DominatorTree>(Function)</tt>. </p>
473 <p>To write a correct <tt>ModulePass</tt> subclass, derive from
474 <tt>ModulePass</tt> and overload the <tt>runOnModule</tt> method with the
475 following signature:</p>
479 <!-- _______________________________________________________________________ -->
480 <div class="doc_subsubsection">
481 <a name="runOnModule">The <tt>runOnModule</tt> method</a>
484 <div class="doc_text">
486 <div class="doc_code"><pre>
487 <b>virtual bool</b> runOnModule(Module &M) = 0;
490 <p>The <tt>runOnModule</tt> method performs the interesting work of the pass.
491 It should return true if the module was modified by the transformation and
496 <!-- ======================================================================= -->
497 <div class="doc_subsection">
498 <a name="CallGraphSCCPass">The <tt>CallGraphSCCPass</tt> class</a>
501 <div class="doc_text">
504 href="http://llvm.org/doxygen/classllvm_1_1CallGraphSCCPass.html">CallGraphSCCPass</a></tt>"
505 is used by passes that need to traverse the program bottom-up on the call graph
506 (callees before callers). Deriving from CallGraphSCCPass provides some
507 mechanics for building and traversing the CallGraph, but also allows the system
508 to optimize execution of CallGraphSCCPass's. If your pass meets the
509 requirements outlined below, and doesn't meet the requirements of a <tt><a
510 href="#FunctionPass">FunctionPass</a></tt> or <tt><a
511 href="#BasicBlockPass">BasicBlockPass</a></tt>, you should derive from
512 <tt>CallGraphSCCPass</tt>.</p>
514 <p><b>TODO</b>: explain briefly what SCC, Tarjan's algo, and B-U mean.</p>
516 <p>To be explicit, <tt>CallGraphSCCPass</tt> subclasses are:</p>
520 <li>... <em>not allowed</em> to modify any <tt>Function</tt>s that are not in
521 the current SCC.</li>
523 <li>... <em>allowed</em> to inspect any Function's other than those in the
524 current SCC and the direct callees of the SCC.</li>
526 <li>... <em>required</em> to preserve the current CallGraph object, updating it
527 to reflect any changes made to the program.</li>
529 <li>... <em>not allowed</em> to add or remove SCC's from the current Module,
530 though they may change the contents of an SCC.</li>
532 <li>... <em>allowed</em> to add or remove global variables from the current
535 <li>... <em>allowed</em> to maintain state across invocations of
536 <a href="#runOnSCC"><tt>runOnSCC</tt></a> (including global data).</li>
539 <p>Implementing a <tt>CallGraphSCCPass</tt> is slightly tricky in some cases
540 because it has to handle SCCs with more than one node in it. All of the virtual
541 methods described below should return true if they modified the program, or
542 false if they didn't.</p>
546 <!-- _______________________________________________________________________ -->
547 <div class="doc_subsubsection">
548 <a name="doInitialization_scc">The <tt>doInitialization(CallGraph &)</tt>
552 <div class="doc_text">
554 <div class="doc_code"><pre>
555 <b>virtual bool</b> doInitialization(CallGraph &CG);
558 <p>The <tt>doIninitialize</tt> method is allowed to do most of the things that
559 <tt>CallGraphSCCPass</tt>'s are not allowed to do. They can add and remove
560 functions, get pointers to functions, etc. The <tt>doInitialization</tt> method
561 is designed to do simple initialization type of stuff that does not depend on
562 the SCCs being processed. The <tt>doInitialization</tt> method call is not
563 scheduled to overlap with any other pass executions (thus it should be very
568 <!-- _______________________________________________________________________ -->
569 <div class="doc_subsubsection">
570 <a name="runOnSCC">The <tt>runOnSCC</tt> method</a>
573 <div class="doc_text">
575 <div class="doc_code"><pre>
576 <b>virtual bool</b> runOnSCC(const std::vector<CallGraphNode *> &SCCM) = 0;
579 <p>The <tt>runOnSCC</tt> method performs the interesting work of the pass, and
580 should return true if the module was modified by the transformation, false
585 <!-- _______________________________________________________________________ -->
586 <div class="doc_subsubsection">
587 <a name="doFinalization_scc">The <tt>doFinalization(CallGraph
588 &)</tt> method</a>
591 <div class="doc_text">
593 <div class="doc_code"><pre>
594 <b>virtual bool</b> doFinalization(CallGraph &CG);
597 <p>The <tt>doFinalization</tt> method is an infrequently used method that is
598 called when the pass framework has finished calling <a
599 href="#runOnFunction"><tt>runOnFunction</tt></a> for every function in the
600 program being compiled.</p>
604 <!-- ======================================================================= -->
605 <div class="doc_subsection">
606 <a name="FunctionPass">The <tt>FunctionPass</tt> class</a>
609 <div class="doc_text">
611 <p>In contrast to <tt>ModulePass</tt> subclasses, <tt><a
612 href="http://llvm.org/doxygen/classllvm_1_1Pass.html">FunctionPass</a></tt>
613 subclasses do have a predictable, local behavior that can be expected by the
614 system. All <tt>FunctionPass</tt> execute on each function in the program
615 independent of all of the other functions in the program.
616 <tt>FunctionPass</tt>'s do not require that they are executed in a particular
617 order, and <tt>FunctionPass</tt>'s do not modify external functions.</p>
619 <p>To be explicit, <tt>FunctionPass</tt> subclasses are not allowed to:</p>
622 <li>Modify a Function other than the one currently being processed.</li>
623 <li>Add or remove Function's from the current Module.</li>
624 <li>Add or remove global variables from the current Module.</li>
625 <li>Maintain state across invocations of
626 <a href="#runOnFunction"><tt>runOnFunction</tt></a> (including global data)</li>
629 <p>Implementing a <tt>FunctionPass</tt> is usually straightforward (See the <a
630 href="#basiccode">Hello World</a> pass for example). <tt>FunctionPass</tt>'s
631 may overload three virtual methods to do their work. All of these methods
632 should return true if they modified the program, or false if they didn't.</p>
636 <!-- _______________________________________________________________________ -->
637 <div class="doc_subsubsection">
638 <a name="doInitialization_mod">The <tt>doInitialization(Module &)</tt>
642 <div class="doc_text">
644 <div class="doc_code"><pre>
645 <b>virtual bool</b> doInitialization(Module &M);
648 <p>The <tt>doIninitialize</tt> method is allowed to do most of the things that
649 <tt>FunctionPass</tt>'s are not allowed to do. They can add and remove
650 functions, get pointers to functions, etc. The <tt>doInitialization</tt> method
651 is designed to do simple initialization type of stuff that does not depend on
652 the functions being processed. The <tt>doInitialization</tt> method call is not
653 scheduled to overlap with any other pass executions (thus it should be very
656 <p>A good example of how this method should be used is the <a
657 href="http://llvm.org/doxygen/LowerAllocations_8cpp-source.html">LowerAllocations</a>
658 pass. This pass converts <tt>malloc</tt> and <tt>free</tt> instructions into
659 platform dependent <tt>malloc()</tt> and <tt>free()</tt> function calls. It
660 uses the <tt>doInitialization</tt> method to get a reference to the malloc and
661 free functions that it needs, adding prototypes to the module if necessary.</p>
665 <!-- _______________________________________________________________________ -->
666 <div class="doc_subsubsection">
667 <a name="runOnFunction">The <tt>runOnFunction</tt> method</a>
670 <div class="doc_text">
672 <div class="doc_code"><pre>
673 <b>virtual bool</b> runOnFunction(Function &F) = 0;
676 <p>The <tt>runOnFunction</tt> method must be implemented by your subclass to do
677 the transformation or analysis work of your pass. As usual, a true value should
678 be returned if the function is modified.</p>
682 <!-- _______________________________________________________________________ -->
683 <div class="doc_subsubsection">
684 <a name="doFinalization_mod">The <tt>doFinalization(Module
685 &)</tt> method</a>
688 <div class="doc_text">
690 <div class="doc_code"><pre>
691 <b>virtual bool</b> doFinalization(Module &M);
694 <p>The <tt>doFinalization</tt> method is an infrequently used method that is
695 called when the pass framework has finished calling <a
696 href="#runOnFunction"><tt>runOnFunction</tt></a> for every function in the
697 program being compiled.</p>
701 <!-- ======================================================================= -->
702 <div class="doc_subsection">
703 <a name="LoopPass">The <tt>LoopPass</tt> class </a>
706 <div class="doc_text">
708 <p> All <tt>LoopPass</tt> execute on each loop in the function independent of
709 all of the other loops in the function. <tt>LoopPass</tt> processes loops in
710 loop nest order such that outer most loop is processed last. </p>
712 <p> <tt>LoopPass</tt> subclasses are allowed to update loop nest using
713 <tt>LPPassManager</tt> interface. Implementing a loop pass is usually
714 straightforward. <tt>Looppass</tt>'s may overload three virtual methods to
715 do their work. All these methods should return true if they modified the
716 program, or false if they didn't. </p>
719 <!-- _______________________________________________________________________ -->
720 <div class="doc_subsubsection">
721 <a name="doInitialization_loop">The <tt>doInitialization(Loop *,
722 LPPassManager &)</tt>
726 <div class="doc_text">
728 <div class="doc_code"><pre>
729 <b>virtual bool</b> doInitialization(Loop *, LPPassManager &LPM);
732 The <tt>doInitialization</tt> method is designed to do simple initialization
733 type of stuff that does not depend on the functions being processed. The
734 <tt>doInitialization</tt> method call is not scheduled to overlap with any
735 other pass executions (thus it should be very fast). LPPassManager
736 interface should be used to access Function or Module level analysis
742 <!-- _______________________________________________________________________ -->
743 <div class="doc_subsubsection">
744 <a name="runOnLoop">The <tt>runOnLoop</tt> method</a>
747 <div class="doc_text">
749 <div class="doc_code"><pre>
750 <b>virtual bool</b> runOnLoop(Loop *, LPPassManager &LPM) = 0;
753 <p>The <tt>runOnLoop</tt> method must be implemented by your subclass to do
754 the transformation or analysis work of your pass. As usual, a true value should
755 be returned if the function is modified. <tt>LPPassManager</tt> interface
756 should be used to update loop nest.</p>
760 <!-- _______________________________________________________________________ -->
761 <div class="doc_subsubsection">
762 <a name="doFinalization_loop">The <tt>doFinalization()</tt> method</a>
765 <div class="doc_text">
767 <div class="doc_code"><pre>
768 <b>virtual bool</b> doFinalization();
771 <p>The <tt>doFinalization</tt> method is an infrequently used method that is
772 called when the pass framework has finished calling <a
773 href="#runOnLoop"><tt>runOnLoop</tt></a> for every loop in the
774 program being compiled. </p>
780 <!-- ======================================================================= -->
781 <div class="doc_subsection">
782 <a name="BasicBlockPass">The <tt>BasicBlockPass</tt> class</a>
785 <div class="doc_text">
787 <p><tt>BasicBlockPass</tt>'s are just like <a
788 href="#FunctionPass"><tt>FunctionPass</tt></a>'s, except that they must limit
789 their scope of inspection and modification to a single basic block at a time.
790 As such, they are <b>not</b> allowed to do any of the following:</p>
793 <li>Modify or inspect any basic blocks outside of the current one</li>
794 <li>Maintain state across invocations of
795 <a href="#runOnBasicBlock"><tt>runOnBasicBlock</tt></a></li>
796 <li>Modify the control flow graph (by altering terminator instructions)</li>
797 <li>Any of the things forbidden for
798 <a href="#FunctionPass"><tt>FunctionPass</tt></a>es.</li>
801 <p><tt>BasicBlockPass</tt>es are useful for traditional local and "peephole"
802 optimizations. They may override the same <a
803 href="#doInitialization_mod"><tt>doInitialization(Module &)</tt></a> and <a
804 href="#doFinalization_mod"><tt>doFinalization(Module &)</tt></a> methods that <a
805 href="#FunctionPass"><tt>FunctionPass</tt></a>'s have, but also have the following virtual methods that may also be implemented:</p>
809 <!-- _______________________________________________________________________ -->
810 <div class="doc_subsubsection">
811 <a name="doInitialization_fn">The <tt>doInitialization(Function
812 &)</tt> method</a>
815 <div class="doc_text">
817 <div class="doc_code"><pre>
818 <b>virtual bool</b> doInitialization(Function &F);
821 <p>The <tt>doIninitialize</tt> method is allowed to do most of the things that
822 <tt>BasicBlockPass</tt>'s are not allowed to do, but that
823 <tt>FunctionPass</tt>'s can. The <tt>doInitialization</tt> method is designed
824 to do simple initialization that does not depend on the
825 BasicBlocks being processed. The <tt>doInitialization</tt> method call is not
826 scheduled to overlap with any other pass executions (thus it should be very
831 <!-- _______________________________________________________________________ -->
832 <div class="doc_subsubsection">
833 <a name="runOnBasicBlock">The <tt>runOnBasicBlock</tt> method</a>
836 <div class="doc_text">
838 <div class="doc_code"><pre>
839 <b>virtual bool</b> runOnBasicBlock(BasicBlock &BB) = 0;
842 <p>Override this function to do the work of the <tt>BasicBlockPass</tt>. This
843 function is not allowed to inspect or modify basic blocks other than the
844 parameter, and are not allowed to modify the CFG. A true value must be returned
845 if the basic block is modified.</p>
849 <!-- _______________________________________________________________________ -->
850 <div class="doc_subsubsection">
851 <a name="doFinalization_fn">The <tt>doFinalization(Function &)</tt>
855 <div class="doc_text">
857 <div class="doc_code"><pre>
858 <b>virtual bool</b> doFinalization(Function &F);
861 <p>The <tt>doFinalization</tt> method is an infrequently used method that is
862 called when the pass framework has finished calling <a
863 href="#runOnBasicBlock"><tt>runOnBasicBlock</tt></a> for every BasicBlock in the
864 program being compiled. This can be used to perform per-function
869 <!-- ======================================================================= -->
870 <div class="doc_subsection">
871 <a name="MachineFunctionPass">The <tt>MachineFunctionPass</tt> class</a>
874 <div class="doc_text">
876 <p>A <tt>MachineFunctionPass</tt> is a part of the LLVM code generator that
877 executes on the machine-dependent representation of each LLVM function in the
878 program. A <tt>MachineFunctionPass</tt> is also a <tt>FunctionPass</tt>, so all
879 the restrictions that apply to a <tt>FunctionPass</tt> also apply to it.
880 <tt>MachineFunctionPass</tt>es also have additional restrictions. In particular,
881 <tt>MachineFunctionPass</tt>es are not allowed to do any of the following:</p>
884 <li>Modify any LLVM Instructions, BasicBlocks or Functions.</li>
885 <li>Modify a MachineFunction other than the one currently being processed.</li>
886 <li>Add or remove MachineFunctions from the current Module.</li>
887 <li>Add or remove global variables from the current Module.</li>
888 <li>Maintain state across invocations of <a
889 href="#runOnMachineFunction"><tt>runOnMachineFunction</tt></a> (including global
895 <!-- _______________________________________________________________________ -->
896 <div class="doc_subsubsection">
897 <a name="runOnMachineFunction">The <tt>runOnMachineFunction(MachineFunction
898 &MF)</tt> method</a>
901 <div class="doc_text">
903 <div class="doc_code"><pre>
904 <b>virtual bool</b> runOnMachineFunction(MachineFunction &MF) = 0;
907 <p><tt>runOnMachineFunction</tt> can be considered the main entry point of a
908 <tt>MachineFunctionPass</tt>; that is, you should override this method to do the
909 work of your <tt>MachineFunctionPass</tt>.</p>
911 <p>The <tt>runOnMachineFunction</tt> method is called on every
912 <tt>MachineFunction</tt> in a <tt>Module</tt>, so that the
913 <tt>MachineFunctionPass</tt> may perform optimizations on the machine-dependent
914 representation of the function. If you want to get at the LLVM <tt>Function</tt>
915 for the <tt>MachineFunction</tt> you're working on, use
916 <tt>MachineFunction</tt>'s <tt>getFunction()</tt> accessor method -- but
917 remember, you may not modify the LLVM <tt>Function</tt> or its contents from a
918 <tt>MachineFunctionPass</tt>.</p>
922 <!-- *********************************************************************** -->
923 <div class="doc_section">
924 <a name="registration">Pass registration</a>
926 <!-- *********************************************************************** -->
928 <div class="doc_text">
930 <p>In the <a href="#basiccode">Hello World</a> example pass we illustrated how
931 pass registration works, and discussed some of the reasons that it is used and
932 what it does. Here we discuss how and why passes are registered.</p>
934 <p>As we saw above, passes are registered with the <b><tt>RegisterPass</tt></b>
935 template, which requires you to pass at least two
936 parameters. The first parameter is the name of the pass that is to be used on
937 the command line to specify that the pass should be added to a program (for
938 example, with <tt>opt</tt> or <tt>bugpoint</tt>). The second argument is the
939 name of the pass, which is to be used for the <tt>--help</tt> output of
941 well as for debug output generated by the <tt>--debug-pass</tt> option.</p>
943 <p>If you want your pass to be easily dumpable, you should
944 implement the virtual <tt>print</tt> method:</p>
948 <!-- _______________________________________________________________________ -->
949 <div class="doc_subsubsection">
950 <a name="print">The <tt>print</tt> method</a>
953 <div class="doc_text">
955 <div class="doc_code"><pre>
956 <b>virtual void</b> print(llvm::OStream &O, <b>const</b> Module *M) <b>const</b>;
959 <p>The <tt>print</tt> method must be implemented by "analyses" in order to print
960 a human readable version of the analysis results. This is useful for debugging
961 an analysis itself, as well as for other people to figure out how an analysis
962 works. Use the <tt>opt -analyze</tt> argument to invoke this method.</p>
964 <p>The <tt>llvm::OStream</tt> parameter specifies the stream to write the results on,
965 and the <tt>Module</tt> parameter gives a pointer to the top level module of the
966 program that has been analyzed. Note however that this pointer may be null in
967 certain circumstances (such as calling the <tt>Pass::dump()</tt> from a
968 debugger), so it should only be used to enhance debug output, it should not be
973 <!-- *********************************************************************** -->
974 <div class="doc_section">
975 <a name="interaction">Specifying interactions between passes</a>
977 <!-- *********************************************************************** -->
979 <div class="doc_text">
981 <p>One of the main responsibilities of the <tt>PassManager</tt> is the make sure
982 that passes interact with each other correctly. Because <tt>PassManager</tt>
983 tries to <a href="#passmanager">optimize the execution of passes</a> it must
984 know how the passes interact with each other and what dependencies exist between
985 the various passes. To track this, each pass can declare the set of passes that
986 are required to be executed before the current pass, and the passes which are
987 invalidated by the current pass.</p>
989 <p>Typically this functionality is used to require that analysis results are
990 computed before your pass is run. Running arbitrary transformation passes can
991 invalidate the computed analysis results, which is what the invalidation set
992 specifies. If a pass does not implement the <tt><a
993 href="#getAnalysisUsage">getAnalysisUsage</a></tt> method, it defaults to not
994 having any prerequisite passes, and invalidating <b>all</b> other passes.</p>
998 <!-- _______________________________________________________________________ -->
999 <div class="doc_subsubsection">
1000 <a name="getAnalysisUsage">The <tt>getAnalysisUsage</tt> method</a>
1003 <div class="doc_text">
1005 <div class="doc_code"><pre>
1006 <b>virtual void</b> getAnalysisUsage(AnalysisUsage &Info) <b>const</b>;
1009 <p>By implementing the <tt>getAnalysisUsage</tt> method, the required and
1010 invalidated sets may be specified for your transformation. The implementation
1011 should fill in the <tt><a
1012 href="http://llvm.org/doxygen/classllvm_1_1AnalysisUsage.html">AnalysisUsage</a></tt>
1013 object with information about which passes are required and not invalidated. To
1014 do this, a pass may call any of the following methods on the AnalysisUsage
1018 <!-- _______________________________________________________________________ -->
1019 <div class="doc_subsubsection">
1020 <a name="AU::addRequired">The <tt>AnalysisUsage::addRequired<></tt> and <tt>AnalysisUsage::addRequiredTransitive<></tt> methods</a>
1023 <div class="doc_text">
1025 If your pass requires a previous pass to be executed (an analysis for example),
1026 it can use one of these methods to arrange for it to be run before your pass.
1027 LLVM has many different types of analyses and passes that can be required,
1028 spanning the range from <tt>DominatorSet</tt> to <tt>BreakCriticalEdges</tt>.
1029 Requiring <tt>BreakCriticalEdges</tt>, for example, guarantees that there will
1030 be no critical edges in the CFG when your pass has been run.
1034 Some analyses chain to other analyses to do their job. For example, an <a
1035 href="AliasAnalysis.html">AliasAnalysis</a> implementation is required to <a
1036 href="AliasAnalysis.html#chaining">chain</a> to other alias analysis passes. In
1037 cases where analyses chain, the <tt>addRequiredTransitive</tt> method should be
1038 used instead of the <tt>addRequired</tt> method. This informs the PassManager
1039 that the transitively required pass should be alive as long as the requiring
1044 <!-- _______________________________________________________________________ -->
1045 <div class="doc_subsubsection">
1046 <a name="AU::addPreserved">The <tt>AnalysisUsage::addPreserved<></tt> method</a>
1049 <div class="doc_text">
1051 One of the jobs of the PassManager is to optimize how and when analyses are run.
1052 In particular, it attempts to avoid recomputing data unless it needs to. For
1053 this reason, passes are allowed to declare that they preserve (i.e., they don't
1054 invalidate) an existing analysis if it's available. For example, a simple
1055 constant folding pass would not modify the CFG, so it can't possibly affect the
1056 results of dominator analysis. By default, all passes are assumed to invalidate
1061 The <tt>AnalysisUsage</tt> class provides several methods which are useful in
1062 certain circumstances that are related to <tt>addPreserved</tt>. In particular,
1063 the <tt>setPreservesAll</tt> method can be called to indicate that the pass does
1064 not modify the LLVM program at all (which is true for analyses), and the
1065 <tt>setPreservesCFG</tt> method can be used by transformations that change
1066 instructions in the program but do not modify the CFG or terminator instructions
1067 (note that this property is implicitly set for <a
1068 href="#BasicBlockPass">BasicBlockPass</a>'s).
1072 <tt>addPreserved</tt> is particularly useful for transformations like
1073 <tt>BreakCriticalEdges</tt>. This pass knows how to update a small set of loop
1074 and dominator related analyses if they exist, so it can preserve them, despite
1075 the fact that it hacks on the CFG.
1079 <!-- _______________________________________________________________________ -->
1080 <div class="doc_subsubsection">
1081 <a name="AU::examples">Example implementations of <tt>getAnalysisUsage</tt></a>
1084 <div class="doc_text">
1086 <div class="doc_code"><pre>
1087 <i>// This is an example implementation from an analysis, which does not modify
1088 // the program at all, yet has a prerequisite.</i>
1089 <b>void</b> <a href="http://llvm.org/doxygen/classllvm_1_1PostDominanceFrontier.html">PostDominanceFrontier</a>::getAnalysisUsage(AnalysisUsage &AU) <b>const</b> {
1090 AU.setPreservesAll();
1091 AU.addRequired<<a href="http://llvm.org/doxygen/classllvm_1_1PostDominatorTree.html">PostDominatorTree</a>>();
1097 <div class="doc_code"><pre>
1098 <i>// This example modifies the program, but does not modify the CFG</i>
1099 <b>void</b> <a href="http://llvm.org/doxygen/structLICM.html">LICM</a>::getAnalysisUsage(AnalysisUsage &AU) <b>const</b> {
1100 AU.setPreservesCFG();
1101 AU.addRequired<<a href="http://llvm.org/doxygen/classllvm_1_1LoopInfo.html">LoopInfo</a>>();
1107 <!-- _______________________________________________________________________ -->
1108 <div class="doc_subsubsection">
1109 <a name="getAnalysis">The <tt>getAnalysis<></tt> and <tt>getAnalysisToUpdate<></tt> methods</a>
1112 <div class="doc_text">
1114 <p>The <tt>Pass::getAnalysis<></tt> method is automatically inherited by
1115 your class, providing you with access to the passes that you declared that you
1116 required with the <a href="#getAnalysisUsage"><tt>getAnalysisUsage</tt></a>
1117 method. It takes a single template argument that specifies which pass class you
1118 want, and returns a reference to that pass. For example:</p>
1120 <div class="doc_code"><pre>
1121 bool LICM::runOnFunction(Function &F) {
1122 LoopInfo &LI = getAnalysis<LoopInfo>();
1127 <p>This method call returns a reference to the pass desired. You may get a
1128 runtime assertion failure if you attempt to get an analysis that you did not
1129 declare as required in your <a
1130 href="#getAnalysisUsage"><tt>getAnalysisUsage</tt></a> implementation. This
1131 method can be called by your <tt>run*</tt> method implementation, or by any
1132 other local method invoked by your <tt>run*</tt> method.
1134 A module level pass can use function level analysis info using this interface.
1137 <div class="doc_code"><pre>
1138 bool ModuleLevelPass::runOnModule(Module &M) {
1140 DominatorTree &DT = getAnalysis<DominatorTree>(Function &F);
1145 In above example, runOnFunction for DominatorTree is called by pass manager
1146 before returning a reference to the desired pass.</p>
1149 If your pass is capable of updating analyses if they exist (e.g.,
1150 <tt>BreakCriticalEdges</tt>, as described above), you can use the
1151 <tt>getAnalysisToUpdate</tt> method, which returns a pointer to the analysis if
1152 it is active. For example:</p>
1154 <div class="doc_code"><pre>
1156 if (DominatorSet *DS = getAnalysisToUpdate<DominatorSet>()) {
1157 <i>// A DominatorSet is active. This code will update it.</i>
1164 <!-- *********************************************************************** -->
1165 <div class="doc_section">
1166 <a name="analysisgroup">Implementing Analysis Groups</a>
1168 <!-- *********************************************************************** -->
1170 <div class="doc_text">
1172 <p>Now that we understand the basics of how passes are defined, how the are
1173 used, and how they are required from other passes, it's time to get a little bit
1174 fancier. All of the pass relationships that we have seen so far are very
1175 simple: one pass depends on one other specific pass to be run before it can run.
1176 For many applications, this is great, for others, more flexibility is
1179 <p>In particular, some analyses are defined such that there is a single simple
1180 interface to the analysis results, but multiple ways of calculating them.
1181 Consider alias analysis for example. The most trivial alias analysis returns
1182 "may alias" for any alias query. The most sophisticated analysis a
1183 flow-sensitive, context-sensitive interprocedural analysis that can take a
1184 significant amount of time to execute (and obviously, there is a lot of room
1185 between these two extremes for other implementations). To cleanly support
1186 situations like this, the LLVM Pass Infrastructure supports the notion of
1187 Analysis Groups.</p>
1191 <!-- _______________________________________________________________________ -->
1192 <div class="doc_subsubsection">
1193 <a name="agconcepts">Analysis Group Concepts</a>
1196 <div class="doc_text">
1198 <p>An Analysis Group is a single simple interface that may be implemented by
1199 multiple different passes. Analysis Groups can be given human readable names
1200 just like passes, but unlike passes, they need not derive from the <tt>Pass</tt>
1201 class. An analysis group may have one or more implementations, one of which is
1202 the "default" implementation.</p>
1204 <p>Analysis groups are used by client passes just like other passes are: the
1205 <tt>AnalysisUsage::addRequired()</tt> and <tt>Pass::getAnalysis()</tt> methods.
1206 In order to resolve this requirement, the <a href="#passmanager">PassManager</a>
1207 scans the available passes to see if any implementations of the analysis group
1208 are available. If none is available, the default implementation is created for
1209 the pass to use. All standard rules for <A href="#interaction">interaction
1210 between passes</a> still apply.</p>
1212 <p>Although <a href="#registration">Pass Registration</a> is optional for normal
1213 passes, all analysis group implementations must be registered, and must use the
1214 <A href="#registerag"><tt>RegisterAnalysisGroup</tt></a> template to join the
1215 implementation pool. Also, a default implementation of the interface
1216 <b>must</b> be registered with <A
1217 href="#registerag"><tt>RegisterAnalysisGroup</tt></a>.</p>
1219 <p>As a concrete example of an Analysis Group in action, consider the <a
1220 href="http://llvm.org/doxygen/classllvm_1_1AliasAnalysis.html">AliasAnalysis</a>
1221 analysis group. The default implementation of the alias analysis interface (the
1223 href="http://llvm.org/doxygen/structBasicAliasAnalysis.html">basicaa</a></tt>
1224 pass) just does a few simple checks that don't require significant analysis to
1225 compute (such as: two different globals can never alias each other, etc).
1226 Passes that use the <tt><a
1227 href="http://llvm.org/doxygen/classllvm_1_1AliasAnalysis.html">AliasAnalysis</a></tt>
1228 interface (for example the <tt><a
1229 href="http://llvm.org/doxygen/structGCSE.html">gcse</a></tt> pass), do
1230 not care which implementation of alias analysis is actually provided, they just
1231 use the designated interface.</p>
1233 <p>From the user's perspective, commands work just like normal. Issuing the
1234 command '<tt>opt -gcse ...</tt>' will cause the <tt>basicaa</tt> class to be
1235 instantiated and added to the pass sequence. Issuing the command '<tt>opt
1236 -somefancyaa -gcse ...</tt>' will cause the <tt>gcse</tt> pass to use the
1237 <tt>somefancyaa</tt> alias analysis (which doesn't actually exist, it's just a
1238 hypothetical example) instead.</p>
1242 <!-- _______________________________________________________________________ -->
1243 <div class="doc_subsubsection">
1244 <a name="registerag">Using <tt>RegisterAnalysisGroup</tt></a>
1247 <div class="doc_text">
1249 <p>The <tt>RegisterAnalysisGroup</tt> template is used to register the analysis
1250 group itself as well as add pass implementations to the analysis group. First,
1251 an analysis should be registered, with a human readable name provided for it.
1252 Unlike registration of passes, there is no command line argument to be specified
1253 for the Analysis Group Interface itself, because it is "abstract":</p>
1255 <div class="doc_code"><pre>
1256 <b>static</b> RegisterAnalysisGroup<<a href="http://llvm.org/doxygen/classllvm_1_1AliasAnalysis.html">AliasAnalysis</a>> A("<i>Alias Analysis</i>");
1259 <p>Once the analysis is registered, passes can declare that they are valid
1260 implementations of the interface by using the following code:</p>
1262 <div class="doc_code"><pre>
1264 //<i> Analysis Group implementations <b>must</b> be registered normally...</i>
1265 RegisterPass<FancyAA>
1266 B("<i>somefancyaa</i>", "<i>A more complex alias analysis implementation</i>");
1268 //<i> Declare that we implement the AliasAnalysis interface</i>
1269 RegisterAnalysisGroup<<a href="http://llvm.org/doxygen/classllvm_1_1AliasAnalysis.html">AliasAnalysis</a>> C(B);
1273 <p>This just shows a class <tt>FancyAA</tt> that is registered normally, then
1274 uses the <tt>RegisterAnalysisGroup</tt> template to "join" the <tt><a
1275 href="http://llvm.org/doxygen/classllvm_1_1AliasAnalysis.html">AliasAnalysis</a></tt>
1276 analysis group. Every implementation of an analysis group should join using
1277 this template. A single pass may join multiple different analysis groups with
1280 <div class="doc_code"><pre>
1282 //<i> Analysis Group implementations <b>must</b> be registered normally...</i>
1283 RegisterPass<<a href="http://llvm.org/doxygen/structBasicAliasAnalysis.html">BasicAliasAnalysis</a>>
1284 D("<i>basicaa</i>", "<i>Basic Alias Analysis (default AA impl)</i>");
1286 //<i> Declare that we implement the AliasAnalysis interface</i>
1287 RegisterAnalysisGroup<<a href="http://llvm.org/doxygen/classllvm_1_1AliasAnalysis.html">AliasAnalysis</a>, <b>true</b>> E(D);
1291 <p>Here we show how the default implementation is specified (using the extra
1292 argument to the <tt>RegisterAnalysisGroup</tt> template). There must be exactly
1293 one default implementation available at all times for an Analysis Group to be
1294 used. Here we declare that the <tt><a
1295 href="http://llvm.org/doxygen/structBasicAliasAnalysis.html">BasicAliasAnalysis</a></tt>
1296 pass is the default implementation for the interface.</p>
1300 <!-- *********************************************************************** -->
1301 <div class="doc_section">
1302 <a name="passStatistics">Pass Statistics</a>
1304 <!-- *********************************************************************** -->
1306 <div class="doc_text">
1308 href="http://llvm.org/doxygen/Statistic_8h-source.html"><tt>Statistic</tt></a>
1309 class is designed to be an easy way to expose various success
1310 metrics from passes. These statistics are printed at the end of a
1311 run, when the -stats command line option is enabled on the command
1312 line. See the <a href="http://llvm.org/docs/ProgrammersManual.html#Statistic">Statistics section</a> in the Programmer's Manual for details.
1317 <!-- *********************************************************************** -->
1318 <div class="doc_section">
1319 <a name="passmanager">What PassManager does</a>
1321 <!-- *********************************************************************** -->
1323 <div class="doc_text">
1326 href="http://llvm.org/doxygen/PassManager_8h-source.html"><tt>PassManager</tt></a>
1328 href="http://llvm.org/doxygen/classllvm_1_1PassManager.html">class</a>
1329 takes a list of passes, ensures their <a href="#interaction">prerequisites</a>
1330 are set up correctly, and then schedules passes to run efficiently. All of the
1331 LLVM tools that run passes use the <tt>PassManager</tt> for execution of these
1334 <p>The <tt>PassManager</tt> does two main things to try to reduce the execution
1335 time of a series of passes:</p>
1338 <li><b>Share analysis results</b> - The PassManager attempts to avoid
1339 recomputing analysis results as much as possible. This means keeping track of
1340 which analyses are available already, which analyses get invalidated, and which
1341 analyses are needed to be run for a pass. An important part of work is that the
1342 <tt>PassManager</tt> tracks the exact lifetime of all analysis results, allowing
1343 it to <a href="#releaseMemory">free memory</a> allocated to holding analysis
1344 results as soon as they are no longer needed.</li>
1346 <li><b>Pipeline the execution of passes on the program</b> - The
1347 <tt>PassManager</tt> attempts to get better cache and memory usage behavior out
1348 of a series of passes by pipelining the passes together. This means that, given
1349 a series of consequtive <a href="#FunctionPass"><tt>FunctionPass</tt></a>'s, it
1350 will execute all of the <a href="#FunctionPass"><tt>FunctionPass</tt></a>'s on
1351 the first function, then all of the <a
1352 href="#FunctionPass"><tt>FunctionPass</tt></a>es on the second function,
1353 etc... until the entire program has been run through the passes.
1355 <p>This improves the cache behavior of the compiler, because it is only touching
1356 the LLVM program representation for a single function at a time, instead of
1357 traversing the entire program. It reduces the memory consumption of compiler,
1358 because, for example, only one <a
1359 href="http://llvm.org/doxygen/classllvm_1_1DominatorSet.html"><tt>DominatorSet</tt></a>
1360 needs to be calculated at a time. This also makes it possible some <a
1361 href="#SMP">interesting enhancements</a> in the future.</p></li>
1365 <p>The effectiveness of the <tt>PassManager</tt> is influenced directly by how
1366 much information it has about the behaviors of the passes it is scheduling. For
1367 example, the "preserved" set is intentionally conservative in the face of an
1368 unimplemented <a href="#getAnalysisUsage"><tt>getAnalysisUsage</tt></a> method.
1369 Not implementing when it should be implemented will have the effect of not
1370 allowing any analysis results to live across the execution of your pass.</p>
1372 <p>The <tt>PassManager</tt> class exposes a <tt>--debug-pass</tt> command line
1373 options that is useful for debugging pass execution, seeing how things work, and
1374 diagnosing when you should be preserving more analyses than you currently are
1375 (To get information about all of the variants of the <tt>--debug-pass</tt>
1376 option, just type '<tt>opt --help-hidden</tt>').</p>
1378 <p>By using the <tt>--debug-pass=Structure</tt> option, for example, we can see
1379 how our <a href="#basiccode">Hello World</a> pass interacts with other passes.
1380 Lets try it out with the <tt>gcse</tt> and <tt>licm</tt> passes:</p>
1382 <div class="doc_code"><pre>
1383 $ opt -load ../../../Debug/lib/Hello.so -gcse -licm --debug-pass=Structure < hello.bc > /dev/null
1385 Function Pass Manager
1386 Dominator Set Construction
1387 Immediate Dominators Construction
1388 Global Common Subexpression Elimination
1389 -- Immediate Dominators Construction
1390 -- Global Common Subexpression Elimination
1391 Natural Loop Construction
1392 Loop Invariant Code Motion
1393 -- Natural Loop Construction
1394 -- Loop Invariant Code Motion
1396 -- Dominator Set Construction
1402 <p>This output shows us when passes are constructed and when the analysis
1403 results are known to be dead (prefixed with '<tt>--</tt>'). Here we see that
1404 GCSE uses dominator and immediate dominator information to do its job. The LICM
1405 pass uses natural loop information, which uses dominator sets, but not immediate
1406 dominators. Because immediate dominators are no longer useful after the GCSE
1407 pass, it is immediately destroyed. The dominator sets are then reused to
1408 compute natural loop information, which is then used by the LICM pass.</p>
1410 <p>After the LICM pass, the module verifier runs (which is automatically added
1411 by the '<tt>opt</tt>' tool), which uses the dominator set to check that the
1412 resultant LLVM code is well formed. After it finishes, the dominator set
1413 information is destroyed, after being computed once, and shared by three
1416 <p>Lets see how this changes when we run the <a href="#basiccode">Hello
1417 World</a> pass in between the two passes:</p>
1419 <div class="doc_code"><pre>
1420 $ opt -load ../../../Debug/lib/Hello.so -gcse -hello -licm --debug-pass=Structure < hello.bc > /dev/null
1422 Function Pass Manager
1423 Dominator Set Construction
1424 Immediate Dominators Construction
1425 Global Common Subexpression Elimination
1426 <b>-- Dominator Set Construction</b>
1427 -- Immediate Dominators Construction
1428 -- Global Common Subexpression Elimination
1429 <b> Hello World Pass
1431 Dominator Set Construction</b>
1432 Natural Loop Construction
1433 Loop Invariant Code Motion
1434 -- Natural Loop Construction
1435 -- Loop Invariant Code Motion
1437 -- Dominator Set Construction
1446 <p>Here we see that the <a href="#basiccode">Hello World</a> pass has killed the
1447 Dominator Set pass, even though it doesn't modify the code at all! To fix this,
1448 we need to add the following <a
1449 href="#getAnalysisUsage"><tt>getAnalysisUsage</tt></a> method to our pass:</p>
1451 <div class="doc_code"><pre>
1452 <i>// We don't modify the program, so we preserve all analyses</i>
1453 <b>virtual void</b> getAnalysisUsage(AnalysisUsage &AU) <b>const</b> {
1454 AU.setPreservesAll();
1458 <p>Now when we run our pass, we get this output:</p>
1460 <div class="doc_code"><pre>
1461 $ opt -load ../../../Debug/lib/Hello.so -gcse -hello -licm --debug-pass=Structure < hello.bc > /dev/null
1462 Pass Arguments: -gcse -hello -licm
1464 Function Pass Manager
1465 Dominator Set Construction
1466 Immediate Dominators Construction
1467 Global Common Subexpression Elimination
1468 -- Immediate Dominators Construction
1469 -- Global Common Subexpression Elimination
1472 Natural Loop Construction
1473 Loop Invariant Code Motion
1474 -- Loop Invariant Code Motion
1475 -- Natural Loop Construction
1477 -- Dominator Set Construction
1486 <p>Which shows that we don't accidentally invalidate dominator information
1487 anymore, and therefore do not have to compute it twice.</p>
1491 <!-- _______________________________________________________________________ -->
1492 <div class="doc_subsubsection">
1493 <a name="releaseMemory">The <tt>releaseMemory</tt> method</a>
1496 <div class="doc_text">
1498 <div class="doc_code"><pre>
1499 <b>virtual void</b> releaseMemory();
1502 <p>The <tt>PassManager</tt> automatically determines when to compute analysis
1503 results, and how long to keep them around for. Because the lifetime of the pass
1504 object itself is effectively the entire duration of the compilation process, we
1505 need some way to free analysis results when they are no longer useful. The
1506 <tt>releaseMemory</tt> virtual method is the way to do this.</p>
1508 <p>If you are writing an analysis or any other pass that retains a significant
1509 amount of state (for use by another pass which "requires" your pass and uses the
1510 <a href="#getAnalysis">getAnalysis</a> method) you should implement
1511 <tt>releaseMEmory</tt> to, well, release the memory allocated to maintain this
1512 internal state. This method is called after the <tt>run*</tt> method for the
1513 class, before the next call of <tt>run*</tt> in your pass.</p>
1517 <!-- *********************************************************************** -->
1518 <div class="doc_section">
1519 <a name="registering">Registering dynamically loaded passes</a>
1521 <!-- *********************************************************************** -->
1523 <div class="doc_text">
1525 <p><i>Size matters</i> when constructing production quality tools using llvm,
1526 both for the purposes of distribution, and for regulating the resident code size
1527 when running on the target system. Therefore, it becomes desirable to
1528 selectively use some passes, while omitting others and maintain the flexibility
1529 to change configurations later on. You want to be able to do all this, and,
1530 provide feedback to the user. This is where pass registration comes into
1533 <p>The fundamental mechanisms for pass registration are the
1534 <tt>MachinePassRegistry</tt> class and subclasses of
1535 <tt>MachinePassRegistryNode</tt>.</p>
1537 <p>An instance of <tt>MachinePassRegistry</tt> is used to maintain a list of
1538 <tt>MachinePassRegistryNode</tt> objects. This instance maintains the list and
1539 communicates additions and deletions to the command line interface.</p>
1541 <p>An instance of <tt>MachinePassRegistryNode</tt> subclass is used to maintain
1542 information provided about a particular pass. This information includes the
1543 command line name, the command help string and the address of the function used
1544 to create an instance of the pass. A global static constructor of one of these
1545 instances <i>registers</i> with a corresponding <tt>MachinePassRegistry</tt>,
1546 the static destructor <i>unregisters</i>. Thus a pass that is statically linked
1547 in the tool will be registered at start up. A dynamically loaded pass will
1548 register on load and unregister at unload.</p>
1552 <!-- _______________________________________________________________________ -->
1553 <div class="doc_subsection">
1554 <a name="registering_existing">Using existing registries</a>
1557 <div class="doc_text">
1559 <p>There are predefined registries to track instruction scheduling
1560 (<tt>RegisterScheduler</tt>) and register allocation (<tt>RegisterRegAlloc</tt>)
1561 machine passes. Here we will describe how to <i>register</i> a register
1562 allocator machine pass.</p>
1564 <p>Implement your register allocator machine pass. In your register allocator
1565 .cpp file add the following include;</p>
1567 <div class="doc_code"><pre>
1568 #include "llvm/CodeGen/RegAllocRegistry.h"
1571 <p>Also in your register allocator .cpp file, define a creator function in the
1574 <div class="doc_code"><pre>
1575 FunctionPass *createMyRegisterAllocator() {
1576 return new MyRegisterAllocator();
1580 <p>Note that the signature of this function should match the type of
1581 <tt>RegisterRegAlloc::FunctionPassCtor</tt>. In the same file add the
1582 "installing" declaration, in the form;</p>
1584 <div class="doc_code"><pre>
1585 static RegisterRegAlloc myRegAlloc("myregalloc",
1586 " my register allocator help string",
1587 createMyRegisterAllocator);
1590 <p>Note the two spaces prior to the help string produces a tidy result on the
1593 <div class="doc_code"><pre>
1596 -regalloc - Register allocator to use: (default = linearscan)
1597 =linearscan - linear scan register allocator
1598 =local - local register allocator
1599 =simple - simple register allocator
1600 =myregalloc - my register allocator help string
1604 <p>And that's it. The user is now free to use <tt>-regalloc=myregalloc</tt> as
1605 an option. Registering instruction schedulers is similar except use the
1606 <tt>RegisterScheduler</tt> class. Note that the
1607 <tt>RegisterScheduler::FunctionPassCtor</tt> is significantly different from
1608 <tt>RegisterRegAlloc::FunctionPassCtor</tt>.</p>
1610 <p>To force the load/linking of your register allocator into the llc/lli tools,
1611 add your creator function's global declaration to "Passes.h" and add a "pseudo"
1612 call line to <tt>llvm/Codegen/LinkAllCodegenComponents.h</tt>.</p>
1617 <!-- _______________________________________________________________________ -->
1618 <div class="doc_subsection">
1619 <a name="registering_new">Creating new registries</a>
1622 <div class="doc_text">
1624 <p>The easiest way to get started is to clone one of the existing registries; we
1625 recommend <tt>llvm/CodeGen/RegAllocRegistry.h</tt>. The key things to modify
1626 are the class name and the <tt>FunctionPassCtor</tt> type.</p>
1628 <p>Then you need to declare the registry. Example: if your pass registry is
1629 <tt>RegisterMyPasses</tt> then define;</p>
1631 <div class="doc_code"><pre>
1632 MachinePassRegistry RegisterMyPasses::Registry;
1635 <p>And finally, declare the command line option for your passes. Example:</p>
1637 <div class="doc_code"><pre>
1638 cl::opt<RegisterMyPasses::FunctionPassCtor, false,
1639 RegisterPassParser<RegisterMyPasses> >
1641 cl::init(&createDefaultMyPass),
1642 cl::desc("my pass option help"));
1645 <p>Here the command option is "mypass", with createDefaultMyPass as the default
1650 <!-- *********************************************************************** -->
1651 <div class="doc_section">
1652 <a name="debughints">Using GDB with dynamically loaded passes</a>
1654 <!-- *********************************************************************** -->
1656 <div class="doc_text">
1658 <p>Unfortunately, using GDB with dynamically loaded passes is not as easy as it
1659 should be. First of all, you can't set a breakpoint in a shared object that has
1660 not been loaded yet, and second of all there are problems with inlined functions
1661 in shared objects. Here are some suggestions to debugging your pass with
1664 <p>For sake of discussion, I'm going to assume that you are debugging a
1665 transformation invoked by <tt>opt</tt>, although nothing described here depends
1670 <!-- _______________________________________________________________________ -->
1671 <div class="doc_subsubsection">
1672 <a name="breakpoint">Setting a breakpoint in your pass</a>
1675 <div class="doc_text">
1677 <p>First thing you do is start <tt>gdb</tt> on the <tt>opt</tt> process:</p>
1679 <div class="doc_code"><pre>
1682 Copyright 2000 Free Software Foundation, Inc.
1683 GDB is free software, covered by the GNU General Public License, and you are
1684 welcome to change it and/or distribute copies of it under certain conditions.
1685 Type "show copying" to see the conditions.
1686 There is absolutely no warranty for GDB. Type "show warranty" for details.
1687 This GDB was configured as "sparc-sun-solaris2.6"...
1691 <p>Note that <tt>opt</tt> has a lot of debugging information in it, so it takes
1692 time to load. Be patient. Since we cannot set a breakpoint in our pass yet
1693 (the shared object isn't loaded until runtime), we must execute the process, and
1694 have it stop before it invokes our pass, but after it has loaded the shared
1695 object. The most foolproof way of doing this is to set a breakpoint in
1696 <tt>PassManager::run</tt> and then run the process with the arguments you
1699 <div class="doc_code"><pre>
1700 (gdb) <b>break llvm::PassManager::run</b>
1701 Breakpoint 1 at 0x2413bc: file Pass.cpp, line 70.
1702 (gdb) <b>run test.bc -load $(LLVMTOP)/llvm/Debug/lib/[libname].so -[passoption]</b>
1703 Starting program: opt test.bc -load $(LLVMTOP)/llvm/Debug/lib/[libname].so -[passoption]
1704 Breakpoint 1, PassManager::run (this=0xffbef174, M=@0x70b298) at Pass.cpp:70
1705 70 bool PassManager::run(Module &M) { return PM->run(M); }
1709 <p>Once the <tt>opt</tt> stops in the <tt>PassManager::run</tt> method you are
1710 now free to set breakpoints in your pass so that you can trace through execution
1711 or do other standard debugging stuff.</p>
1715 <!-- _______________________________________________________________________ -->
1716 <div class="doc_subsubsection">
1717 <a name="debugmisc">Miscellaneous Problems</a>
1720 <div class="doc_text">
1722 <p>Once you have the basics down, there are a couple of problems that GDB has,
1723 some with solutions, some without.</p>
1726 <li>Inline functions have bogus stack information. In general, GDB does a
1727 pretty good job getting stack traces and stepping through inline functions.
1728 When a pass is dynamically loaded however, it somehow completely loses this
1729 capability. The only solution I know of is to de-inline a function (move it
1730 from the body of a class to a .cpp file).</li>
1732 <li>Restarting the program breaks breakpoints. After following the information
1733 above, you have succeeded in getting some breakpoints planted in your pass. Nex
1734 thing you know, you restart the program (i.e., you type '<tt>run</tt>' again),
1735 and you start getting errors about breakpoints being unsettable. The only way I
1736 have found to "fix" this problem is to <tt>delete</tt> the breakpoints that are
1737 already set in your pass, run the program, and re-set the breakpoints once
1738 execution stops in <tt>PassManager::run</tt>.</li>
1742 <p>Hopefully these tips will help with common case debugging situations. If
1743 you'd like to contribute some tips of your own, just contact <a
1744 href="mailto:sabre@nondot.org">Chris</a>.</p>
1748 <!-- *********************************************************************** -->
1749 <div class="doc_section">
1750 <a name="future">Future extensions planned</a>
1752 <!-- *********************************************************************** -->
1754 <div class="doc_text">
1756 <p>Although the LLVM Pass Infrastructure is very capable as it stands, and does
1757 some nifty stuff, there are things we'd like to add in the future. Here is
1758 where we are going:</p>
1762 <!-- _______________________________________________________________________ -->
1763 <div class="doc_subsubsection">
1764 <a name="SMP">Multithreaded LLVM</a>
1767 <div class="doc_text">
1769 <p>Multiple CPU machines are becoming more common and compilation can never be
1770 fast enough: obviously we should allow for a multithreaded compiler. Because of
1771 the semantics defined for passes above (specifically they cannot maintain state
1772 across invocations of their <tt>run*</tt> methods), a nice clean way to
1773 implement a multithreaded compiler would be for the <tt>PassManager</tt> class
1774 to create multiple instances of each pass object, and allow the separate
1775 instances to be hacking on different parts of the program at the same time.</p>
1777 <p>This implementation would prevent each of the passes from having to implement
1778 multithreaded constructs, requiring only the LLVM core to have locking in a few
1779 places (for global resources). Although this is a simple extension, we simply
1780 haven't had time (or multiprocessor machines, thus a reason) to implement this.
1781 Despite that, we have kept the LLVM passes SMP ready, and you should too.</p>
1785 <!-- _______________________________________________________________________ -->
1786 <div class="doc_subsubsection">
1787 <a name="PassFunctionPass"><tt>ModulePass</tt>es requiring <tt>FunctionPass</tt>es</a>
1790 <div class="doc_text">
1792 <p>Currently it is illegal for a <a href="#ModulePass"><tt>ModulePass</tt></a>
1793 to require a <a href="#FunctionPass"><tt>FunctionPass</tt></a>. This is because
1794 there is only one instance of the <a
1795 href="#FunctionPass"><tt>FunctionPass</tt></a> object ever created, thus nowhere
1796 to store information for all of the functions in the program at the same time.
1797 Although this has come up a couple of times before, this has always been worked
1798 around by factoring one big complicated pass into a global and an
1799 interprocedural part, both of which are distinct. In the future, it would be
1800 nice to have this though.</p>
1802 <p>Note that it is no problem for a <a
1803 href="#FunctionPass"><tt>FunctionPass</tt></a> to require the results of a <a
1804 href="#ModulePass"><tt>ModulePass</tt></a>, only the other way around.</p>
1808 <!-- *********************************************************************** -->
1811 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
1812 src="http://jigsaw.w3.org/css-validator/images/vcss" alt="Valid CSS!"></a>
1813 <a href="http://validator.w3.org/check/referer"><img
1814 src="http://www.w3.org/Icons/valid-html401" alt="Valid HTML 4.01!" /></a>
1816 <a href="mailto:sabre@nondot.org">Chris Lattner</a><br>
1817 <a href="http://llvm.org">The LLVM Compiler Infrastructure</a><br>
1818 Last modified: $Date$