| blob | f715a961a0f5f5d7a0c93443aadbf15735e2ae76 |
3 <html>
4 <head>
8 </head>
9 <body>
12 Writing an LLVM Pass
13 </div>
15 <ol>
18 <ul>
22 </ul></li>
24 <ul>
27 <ul>
29 </ul></li>
31 <ul>
37 </ul></li>
39 <ul>
45 </ul></li>
47 <ul>
53 </ul></li>
55 <ul>
59 method</a></li>
62 </ul></li>
64 class</a>
65 <ul>
68 </ul></li>
69 </ul>
71 <ul>
73 </ul></li>
75 <ul>
77 method</a></li>
78 <li><a href="#AU::addRequired">The <tt>AnalysisUsage::addRequired<></tt> and <tt>AnalysisUsage::addRequiredTransitive<></tt> methods</a></li>
83 </ul></li>
85 <ul>
88 </ul></li>
91 <ul>
93 </ul></li>
95 <ul>
98 </ul></li>
100 <ul>
103 </ul></li>
105 <ul>
107 </ul></li>
108 </ol>
113 </div>
115 <!-- *********************************************************************** -->
118 </div>
119 <!-- *********************************************************************** -->
123 <p>The LLVM Pass Framework is an important part of the LLVM system, because LLVM
124 passes are where most of the interesting parts of the compiler exist. Passes
125 perform the transformations and optimizations that make up the compiler, they
126 build the analysis results that are used by these transformations, and they are,
127 above all, a structuring technique for compiler code.</p>
131 class, which implement functionality by overriding virtual methods inherited
138 more information about what your pass does, and how it can be combined with
139 other passes. One of the main features of the LLVM Pass Framework is that it
140 schedules passes to run in an efficient way based on the constraints that your
141 pass meets (which are indicated by which class they derive from).</p>
143 <p>We start by showing you how to construct a pass, everything from setting up
144 the code, to compiling, loading, and executing it. After the basics are down,
145 more advanced features are discussed.</p>
147 </div>
149 <!-- *********************************************************************** -->
152 </div>
153 <!-- *********************************************************************** -->
158 is designed to simply print out the name of non-external functions that exist in
159 the program being compiled. It does not modify the program at all, it just
160 inspects it. The source code and files for this pass are available in the LLVM
163 </div>
165 <!-- ======================================================================= -->
168 </div>
172 <p>First, you need to create a new directory somewhere in the LLVM source
173 base. For this example, we'll assume that you made
175 (Makefile) that will compile the source code for the new pass. To do this,
177 <hr/>
180 # Makefile for hello pass
182 # Path to top level of LLVM heirarchy
183 LEVEL = ../../..
185 # Name of the library to build
186 LIBRARYNAME = Hello
188 # Make the shared library become a loadable module so the tools can
189 # dlopen/dlsym on the resulting library.
190 LOADABLE_MODULE = 1
192 # Tell the build system which LLVM libraries your pass needs. You'll probably
193 # need at least LLVMSystem.a, LLVMSupport.a, LLVMCore.a but possibly several
194 # others too.
195 LLVMLIBS = LLVMCore.a LLVMSupport.a LLVMSystem.a
197 # Include the makefile implementation stuff
198 include $(LEVEL)/Makefile.common
199 </pre></div>
202 directory are to be compiled and linked together into a
205 If your operating system uses a suffix other than .so (such as windows or
206 Mac OS/X), the appropriate extension will be used.</p>
208 <p>Now that we have the build scripts set up, we just need to write the code for
209 the pass itself.</p>
211 </div>
213 <!-- ======================================================================= -->
216 </div>
220 <p>Now that we have a way to compile our new pass, we just have to write it.
221 Start out with:</p>
226 <b>#include</b> "<a href="http://llvm.org/doxygen/raw__ostream_8h.html">llvm/Support/raw_ostream.h</a>"
227 </pre></div>
231 we are operating on <tt><a
233 and we will be doing some printing.</p>
238 </pre></div>
239 <p>... which is required because the functions from the include files
240 live in the llvm namespace.
241 </p>
247 </pre></div>
249 <p>... which starts out an anonymous namespace. Anonymous namespaces are to C++
250 what the "<tt>static</tt>" keyword is to C (at global scope). It makes the
251 things declared inside of the anonymous namespace only visible to the current
252 file. If you're not familiar with them, consult a decent C++ book for more
253 information.</p>
259 </pre></div><p>
263 The different builtin pass subclasses are described in detail <a
266 time.</p>
269 static char ID;
270 Hello() : FunctionPass(&ID) {}
271 </pre></div><p>
273 <p> This declares pass identifier used by LLVM to identify pass. This allows LLVM to
274 avoid using expensive C++ runtime information.</p>
280 }
282 </pre></div>
285 which overloads an abstract virtual method inherited from <a
287 to do our thing, so we just print out our message with the name of each
288 function.</p>
291 char Hello::ID = 0;
292 </pre></div>
294 <p> We initialize pass ID here. LLVM uses ID's address to identify pass so
295 initialization value is not important.</p>
299 false /* Only looks at CFG */,
300 false /* Analysis Pass */);
302 </pre></div>
305 giving it a command line
307 Last two RegisterPass arguments are optional. Their default value is false.
308 If a pass walks CFG without modifying it then third argument is set to true.
309 If a pass is an analysis pass, for example dominator tree pass, then true
310 is supplied as fourth argument. </p>
317 <b>#include</b> "<a href="http://llvm.org/doxygen/raw__ostream_8h.html">llvm/Support/raw_ostream.h</a>"
324 static char ID;
325 Hello() : FunctionPass(&ID) {}
330 }
331 };
333 char Hello::ID = 0;
335 }
336 </pre></div>
339 command in the local directory and you should get a new
340 "<tt>Debug/lib/Hello.so</tt> file. Note that everything in this file is
341 contained in an anonymous namespace: this reflects the fact that passes are self
342 contained units that do not need external interfaces (although they can have
343 them) to be useful.</p>
345 </div>
347 <!-- ======================================================================= -->
350 </div>
354 <p>Now that you have a brand new shiny shared object file, we can use the
355 <tt>opt</tt> command to run an LLVM program through your pass. Because you
356 registered your pass with the <tt>RegisterPass</tt> template, you will be able to
357 use the <tt>opt</tt> tool to access it, once loaded.</p>
359 <p>To test it, follow the example at the end of the <a
361 LLVM. We can now run the bitcode file (<tt>hello.bc</tt>) for the program
362 through our transformation like this (or course, any bitcode file will
363 work):</p>
366 $ opt -load ../../../Debug/lib/Hello.so -hello < hello.bc > /dev/null
367 Hello: __main
368 Hello: puts
369 Hello: main
370 </pre></div>
372 <p>The '<tt>-load</tt>' option specifies that '<tt>opt</tt>' should load your
373 pass as a shared object, which makes '<tt>-hello</tt>' a valid command line
375 pass</a>). Because the hello pass does not modify the program in any
376 interesting way, we just throw away the result of <tt>opt</tt> (sending it to
377 <tt>/dev/null</tt>).</p>
379 <p>To see what happened to the other string you registered, try running
380 <tt>opt</tt> with the <tt>--help</tt> option:</p>
383 $ opt -load ../../../Debug/lib/Hello.so --help
384 OVERVIEW: llvm .bc -> .bc modular optimizer
386 USAGE: opt [options] <input bitcode>
388 OPTIONS:
389 Optimizations available:
390 ...
391 -funcresolve - Resolve Functions
392 -gcse - Global Common Subexpression Elimination
393 -globaldce - Dead Global Elimination
394 <b>-hello - Hello World Pass</b>
395 -indvars - Canonicalize Induction Variables
396 -inline - Function Integration/Inlining
397 -instcombine - Combine redundant instructions
398 ...
399 </pre></div>
401 <p>The pass name get added as the information string for your pass, giving some
402 documentation to users of <tt>opt</tt>. Now that you have a working pass, you
403 would go ahead and make it do the cool transformations you want. Once you get
404 it all working and tested, it may become useful to find out how fast your pass
406 line option (<tt>--time-passes</tt>) that allows you to get information about
407 the execution time of your pass along with the other passes you queue up. For
408 example:</p>
411 $ opt -load ../../../Debug/lib/Hello.so -hello -time-passes < hello.bc > /dev/null
412 Hello: __main
413 Hello: puts
414 Hello: main
415 ===============================================================================
416 ... Pass execution timing report ...
417 ===============================================================================
418 Total Execution Time: 0.02 seconds (0.0479059 wall clock)
420 ---User Time--- --System Time-- --User+System-- ---Wall Time--- --- Pass Name ---
421 0.0100 (100.0%) 0.0000 ( 0.0%) 0.0100 ( 50.0%) 0.0402 ( 84.0%) Bitcode Writer
422 0.0000 ( 0.0%) 0.0100 (100.0%) 0.0100 ( 50.0%) 0.0031 ( 6.4%) Dominator Set Construction
423 0.0000 ( 0.0%) 0.0000 ( 0.0%) 0.0000 ( 0.0%) 0.0013 ( 2.7%) Module Verifier
424 <b> 0.0000 ( 0.0%) 0.0000 ( 0.0%) 0.0000 ( 0.0%) 0.0033 ( 6.9%) Hello World Pass</b>
425 0.0100 (100.0%) 0.0100 (100.0%) 0.0200 (100.0%) 0.0479 (100.0%) TOTAL
426 </pre></div>
428 <p>As you can see, our implementation above is pretty fast :). The additional
429 passes listed are automatically inserted by the '<tt>opt</tt>' tool to verify
430 that the LLVM emitted by your pass is still valid and well formed LLVM, which
431 hasn't been broken somehow.</p>
433 <p>Now that you have seen the basics of the mechanics behind passes, we can talk
434 about some more details of how they work and how to use them.</p>
436 </div>
438 <!-- *********************************************************************** -->
441 </div>
442 <!-- *********************************************************************** -->
446 <p>One of the first things that you should do when designing a new pass is to
447 decide what class you should subclass for your pass. The <a
450 did not discuss why or when this should occur. Here we talk about the classes
451 available, from the most general to the most specific.</p>
453 <p>When choosing a superclass for your Pass, you should choose the <b>most
454 specific</b> class possible, while still being able to meet the requirements
455 listed. This gives the LLVM Pass Infrastructure information necessary to
456 optimize how passes are run, so that the resultant compiler isn't unneccesarily
457 slow.</p>
459 </div>
461 <!-- ======================================================================= -->
464 </div>
470 class. This pass type is used for passes that do not have to be run, do not
471 change state, and never need to be updated. This is not a normal type of
472 transformation or analysis, but can provide information about the current
473 compiler configuration.</p>
475 <p>Although this pass class is very infrequently used, it is important for
476 providing information about the current target machine being compiled for, and
477 other static information that can affect the various transformations.</p>
479 <p><tt>ImmutablePass</tt>es never invalidate other transformations, are never
482 </div>
484 <!-- ======================================================================= -->
487 </div>
493 class is the most general of all superclasses that you can use. Deriving from
494 <tt>ModulePass</tt> indicates that your pass uses the entire program as a unit,
495 refering to function bodies in no predictable order, or adding and removing
496 functions. Because nothing is known about the behavior of <tt>ModulePass</tt>
497 subclasses, no optimization can be done for their execution.</p>
499 <p>A module pass can use function level passes (e.g. dominators) using
500 the getAnalysis interface
501 <tt>getAnalysis<DominatorTree>(llvm::Function *)</tt> to provide the
502 function to retrieve analysis result for, if the function pass does not require
503 any module or immutable passes. Note that this can only be done for functions for which the
504 analysis ran, e.g. in the case of dominators you should only ask for the
505 DominatorTree for function definitions, not declarations.</p>
507 <p>To write a correct <tt>ModulePass</tt> subclass, derive from
508 <tt>ModulePass</tt> and overload the <tt>runOnModule</tt> method with the
509 following signature:</p>
511 </div>
513 <!-- _______________________________________________________________________ -->
516 </div>
521 <b>virtual bool</b> runOnModule(Module &M) = 0;
522 </pre></div>
524 <p>The <tt>runOnModule</tt> method performs the interesting work of the pass.
525 It should return true if the module was modified by the transformation and
526 false otherwise.</p>
528 </div>
530 <!-- ======================================================================= -->
533 </div>
539 is used by passes that need to traverse the program bottom-up on the call graph
540 (callees before callers). Deriving from CallGraphSCCPass provides some
541 mechanics for building and traversing the CallGraph, but also allows the system
542 to optimize execution of CallGraphSCCPass's. If your pass meets the
543 requirements outlined below, and doesn't meet the requirements of a <tt><a
546 <tt>CallGraphSCCPass</tt>.</p>
548 <p><b>TODO</b>: explain briefly what SCC, Tarjan's algo, and B-U mean.</p>
550 <p>To be explicit, <tt>CallGraphSCCPass</tt> subclasses are:</p>
552 <ol>
554 <li>... <em>not allowed</em> to modify any <tt>Function</tt>s that are not in
555 the current SCC.</li>
557 <li>... <em>not allowed</em> to inspect any Function's other than those in the
558 current SCC and the direct callees of the SCC.</li>
560 <li>... <em>required</em> to preserve the current CallGraph object, updating it
561 to reflect any changes made to the program.</li>
563 <li>... <em>not allowed</em> to add or remove SCC's from the current Module,
564 though they may change the contents of an SCC.</li>
566 <li>... <em>allowed</em> to add or remove global variables from the current
567 Module.</li>
569 <li>... <em>allowed</em> to maintain state across invocations of
571 </ol>
573 <p>Implementing a <tt>CallGraphSCCPass</tt> is slightly tricky in some cases
574 because it has to handle SCCs with more than one node in it. All of the virtual
575 methods described below should return true if they modified the program, or
576 false if they didn't.</p>
578 </div>
580 <!-- _______________________________________________________________________ -->
583 method</a>
584 </div>
589 <b>virtual bool</b> doInitialization(CallGraph &CG);
590 </pre></div>
592 <p>The <tt>doIninitialize</tt> method is allowed to do most of the things that
593 <tt>CallGraphSCCPass</tt>'s are not allowed to do. They can add and remove
594 functions, get pointers to functions, etc. The <tt>doInitialization</tt> method
595 is designed to do simple initialization type of stuff that does not depend on
596 the SCCs being processed. The <tt>doInitialization</tt> method call is not
597 scheduled to overlap with any other pass executions (thus it should be very
598 fast).</p>
600 </div>
602 <!-- _______________________________________________________________________ -->
605 </div>
610 <b>virtual bool</b> runOnSCC(const std::vector<CallGraphNode *> &SCCM) = 0;
611 </pre></div>
613 <p>The <tt>runOnSCC</tt> method performs the interesting work of the pass, and
614 should return true if the module was modified by the transformation, false
615 otherwise.</p>
617 </div>
619 <!-- _______________________________________________________________________ -->
622 &)</tt> method</a>
623 </div>
628 <b>virtual bool</b> doFinalization(CallGraph &CG);
629 </pre></div>
631 <p>The <tt>doFinalization</tt> method is an infrequently used method that is
632 called when the pass framework has finished calling <a
634 program being compiled.</p>
636 </div>
638 <!-- ======================================================================= -->
641 </div>
645 <p>In contrast to <tt>ModulePass</tt> subclasses, <tt><a
647 subclasses do have a predictable, local behavior that can be expected by the
648 system. All <tt>FunctionPass</tt> execute on each function in the program
649 independent of all of the other functions in the program.
650 <tt>FunctionPass</tt>'s do not require that they are executed in a particular
651 order, and <tt>FunctionPass</tt>'s do not modify external functions.</p>
653 <p>To be explicit, <tt>FunctionPass</tt> subclasses are not allowed to:</p>
655 <ol>
656 <li>Modify a Function other than the one currently being processed.</li>
657 <li>Add or remove Function's from the current Module.</li>
658 <li>Add or remove global variables from the current Module.</li>
659 <li>Maintain state across invocations of
661 </ol>
663 <p>Implementing a <tt>FunctionPass</tt> is usually straightforward (See the <a
665 may overload three virtual methods to do their work. All of these methods
666 should return true if they modified the program, or false if they didn't.</p>
668 </div>
670 <!-- _______________________________________________________________________ -->
673 method</a>
674 </div>
679 <b>virtual bool</b> doInitialization(Module &M);
680 </pre></div>
682 <p>The <tt>doIninitialize</tt> method is allowed to do most of the things that
683 <tt>FunctionPass</tt>'s are not allowed to do. They can add and remove
684 functions, get pointers to functions, etc. The <tt>doInitialization</tt> method
685 is designed to do simple initialization type of stuff that does not depend on
686 the functions being processed. The <tt>doInitialization</tt> method call is not
687 scheduled to overlap with any other pass executions (thus it should be very
688 fast).</p>
690 <p>A good example of how this method should be used is the <a
692 pass. This pass converts <tt>malloc</tt> and <tt>free</tt> instructions into
693 platform dependent <tt>malloc()</tt> and <tt>free()</tt> function calls. It
694 uses the <tt>doInitialization</tt> method to get a reference to the malloc and
695 free functions that it needs, adding prototypes to the module if necessary.</p>
697 </div>
699 <!-- _______________________________________________________________________ -->
702 </div>
707 <b>virtual bool</b> runOnFunction(Function &F) = 0;
708 </pre></div><p>
710 <p>The <tt>runOnFunction</tt> method must be implemented by your subclass to do
711 the transformation or analysis work of your pass. As usual, a true value should
712 be returned if the function is modified.</p>
714 </div>
716 <!-- _______________________________________________________________________ -->
719 &)</tt> method</a>
720 </div>
725 <b>virtual bool</b> doFinalization(Module &M);
726 </pre></div>
728 <p>The <tt>doFinalization</tt> method is an infrequently used method that is
729 called when the pass framework has finished calling <a
731 program being compiled.</p>
733 </div>
735 <!-- ======================================================================= -->
738 </div>
742 <p> All <tt>LoopPass</tt> execute on each loop in the function independent of
743 all of the other loops in the function. <tt>LoopPass</tt> processes loops in
744 loop nest order such that outer most loop is processed last. </p>
746 <p> <tt>LoopPass</tt> subclasses are allowed to update loop nest using
747 <tt>LPPassManager</tt> interface. Implementing a loop pass is usually
748 straightforward. <tt>Looppass</tt>'s may overload three virtual methods to
749 do their work. All these methods should return true if they modified the
750 program, or false if they didn't. </p>
751 </div>
753 <!-- _______________________________________________________________________ -->
756 LPPassManager &)</tt>
757 method</a>
758 </div>
763 <b>virtual bool</b> doInitialization(Loop *, LPPassManager &LPM);
764 </pre></div>
766 <p>The <tt>doInitialization</tt> method is designed to do simple initialization
767 type of stuff that does not depend on the functions being processed. The
768 <tt>doInitialization</tt> method call is not scheduled to overlap with any
769 other pass executions (thus it should be very fast). LPPassManager
770 interface should be used to access Function or Module level analysis
771 information.</p>
773 </div>
776 <!-- _______________________________________________________________________ -->
779 </div>
784 <b>virtual bool</b> runOnLoop(Loop *, LPPassManager &LPM) = 0;
785 </pre></div><p>
787 <p>The <tt>runOnLoop</tt> method must be implemented by your subclass to do
788 the transformation or analysis work of your pass. As usual, a true value should
789 be returned if the function is modified. <tt>LPPassManager</tt> interface
790 should be used to update loop nest.</p>
792 </div>
794 <!-- _______________________________________________________________________ -->
797 </div>
802 <b>virtual bool</b> doFinalization();
803 </pre></div>
805 <p>The <tt>doFinalization</tt> method is an infrequently used method that is
806 called when the pass framework has finished calling <a
808 program being compiled. </p>
810 </div>
814 <!-- ======================================================================= -->
817 </div>
821 <p><tt>BasicBlockPass</tt>'s are just like <a
823 their scope of inspection and modification to a single basic block at a time.
824 As such, they are <b>not</b> allowed to do any of the following:</p>
826 <ol>
827 <li>Modify or inspect any basic blocks outside of the current one</li>
828 <li>Maintain state across invocations of
830 <li>Modify the control flow graph (by altering terminator instructions)</li>
831 <li>Any of the things forbidden for
833 </ol>
836 optimizations. They may override the same <a
839 href="#FunctionPass"><tt>FunctionPass</tt></a>'s have, but also have the following virtual methods that may also be implemented:</p>
841 </div>
843 <!-- _______________________________________________________________________ -->
846 &)</tt> method</a>
847 </div>
852 <b>virtual bool</b> doInitialization(Function &F);
853 </pre></div>
855 <p>The <tt>doIninitialize</tt> method is allowed to do most of the things that
856 <tt>BasicBlockPass</tt>'s are not allowed to do, but that
857 <tt>FunctionPass</tt>'s can. The <tt>doInitialization</tt> method is designed
858 to do simple initialization that does not depend on the
859 BasicBlocks being processed. The <tt>doInitialization</tt> method call is not
860 scheduled to overlap with any other pass executions (thus it should be very
861 fast).</p>
863 </div>
865 <!-- _______________________________________________________________________ -->
868 </div>
873 <b>virtual bool</b> runOnBasicBlock(BasicBlock &BB) = 0;
874 </pre></div>
876 <p>Override this function to do the work of the <tt>BasicBlockPass</tt>. This
877 function is not allowed to inspect or modify basic blocks other than the
878 parameter, and are not allowed to modify the CFG. A true value must be returned
879 if the basic block is modified.</p>
881 </div>
883 <!-- _______________________________________________________________________ -->
886 method</a>
887 </div>
892 <b>virtual bool</b> doFinalization(Function &F);
893 </pre></div>
895 <p>The <tt>doFinalization</tt> method is an infrequently used method that is
896 called when the pass framework has finished calling <a
898 program being compiled. This can be used to perform per-function
899 finalization.</p>
901 </div>
903 <!-- ======================================================================= -->
906 </div>
910 <p>A <tt>MachineFunctionPass</tt> is a part of the LLVM code generator that
911 executes on the machine-dependent representation of each LLVM function in the
912 program. A <tt>MachineFunctionPass</tt> is also a <tt>FunctionPass</tt>, so all
913 the restrictions that apply to a <tt>FunctionPass</tt> also apply to it.
914 <tt>MachineFunctionPass</tt>es also have additional restrictions. In particular,
915 <tt>MachineFunctionPass</tt>es are not allowed to do any of the following:</p>
917 <ol>
918 <li>Modify any LLVM Instructions, BasicBlocks or Functions.</li>
919 <li>Modify a MachineFunction other than the one currently being processed.</li>
920 <li>Add or remove MachineFunctions from the current Module.</li>
921 <li>Add or remove global variables from the current Module.</li>
922 <li>Maintain state across invocations of <a
924 data)</li>
925 </ol>
927 </div>
929 <!-- _______________________________________________________________________ -->
932 &MF)</tt> method</a>
933 </div>
938 <b>virtual bool</b> runOnMachineFunction(MachineFunction &MF) = 0;
939 </pre></div>
941 <p><tt>runOnMachineFunction</tt> can be considered the main entry point of a
942 <tt>MachineFunctionPass</tt>; that is, you should override this method to do the
943 work of your <tt>MachineFunctionPass</tt>.</p>
945 <p>The <tt>runOnMachineFunction</tt> method is called on every
946 <tt>MachineFunction</tt> in a <tt>Module</tt>, so that the
947 <tt>MachineFunctionPass</tt> may perform optimizations on the machine-dependent
948 representation of the function. If you want to get at the LLVM <tt>Function</tt>
949 for the <tt>MachineFunction</tt> you're working on, use
950 <tt>MachineFunction</tt>'s <tt>getFunction()</tt> accessor method -- but
951 remember, you may not modify the LLVM <tt>Function</tt> or its contents from a
952 <tt>MachineFunctionPass</tt>.</p>
954 </div>
956 <!-- *********************************************************************** -->
959 </div>
960 <!-- *********************************************************************** -->
965 pass registration works, and discussed some of the reasons that it is used and
966 what it does. Here we discuss how and why passes are registered.</p>
968 <p>As we saw above, passes are registered with the <b><tt>RegisterPass</tt></b>
969 template, which requires you to pass at least two
970 parameters. The first parameter is the name of the pass that is to be used on
971 the command line to specify that the pass should be added to a program (for
972 example, with <tt>opt</tt> or <tt>bugpoint</tt>). The second argument is the
973 name of the pass, which is to be used for the <tt>--help</tt> output of
974 programs, as
975 well as for debug output generated by the <tt>--debug-pass</tt> option.</p>
977 <p>If you want your pass to be easily dumpable, you should
978 implement the virtual <tt>print</tt> method:</p>
980 </div>
982 <!-- _______________________________________________________________________ -->
985 </div>
990 <b>virtual void</b> print(std::ostream &O, <b>const</b> Module *M) <b>const</b>;
991 </pre></div>
994 a human readable version of the analysis results. This is useful for debugging
995 an analysis itself, as well as for other people to figure out how an analysis
996 works. Use the <tt>opt -analyze</tt> argument to invoke this method.</p>
998 <p>The <tt>llvm::OStream</tt> parameter specifies the stream to write the results on,
999 and the <tt>Module</tt> parameter gives a pointer to the top level module of the
1000 program that has been analyzed. Note however that this pointer may be null in
1001 certain circumstances (such as calling the <tt>Pass::dump()</tt> from a
1002 debugger), so it should only be used to enhance debug output, it should not be
1003 depended on.</p>
1005 </div>
1007 <!-- *********************************************************************** -->
1010 </div>
1011 <!-- *********************************************************************** -->
1015 <p>One of the main responsibilities of the <tt>PassManager</tt> is to make sure
1016 that passes interact with each other correctly. Because <tt>PassManager</tt>
1018 know how the passes interact with each other and what dependencies exist between
1019 the various passes. To track this, each pass can declare the set of passes that
1020 are required to be executed before the current pass, and the passes which are
1021 invalidated by the current pass.</p>
1023 <p>Typically this functionality is used to require that analysis results are
1024 computed before your pass is run. Running arbitrary transformation passes can
1025 invalidate the computed analysis results, which is what the invalidation set
1026 specifies. If a pass does not implement the <tt><a
1028 having any prerequisite passes, and invalidating <b>all</b> other passes.</p>
1030 </div>
1032 <!-- _______________________________________________________________________ -->
1035 </div>
1040 <b>virtual void</b> getAnalysisUsage(AnalysisUsage &Info) <b>const</b>;
1041 </pre></div>
1043 <p>By implementing the <tt>getAnalysisUsage</tt> method, the required and
1044 invalidated sets may be specified for your transformation. The implementation
1045 should fill in the <tt><a
1047 object with information about which passes are required and not invalidated. To
1048 do this, a pass may call any of the following methods on the AnalysisUsage
1049 object:</p>
1050 </div>
1052 <!-- _______________________________________________________________________ -->
1054 <a name="AU::addRequired">The <tt>AnalysisUsage::addRequired<></tt> and <tt>AnalysisUsage::addRequiredTransitive<></tt> methods</a>
1055 </div>
1058 <p>
1059 If your pass requires a previous pass to be executed (an analysis for example),
1060 it can use one of these methods to arrange for it to be run before your pass.
1061 LLVM has many different types of analyses and passes that can be required,
1062 spanning the range from <tt>DominatorSet</tt> to <tt>BreakCriticalEdges</tt>.
1063 Requiring <tt>BreakCriticalEdges</tt>, for example, guarantees that there will
1064 be no critical edges in the CFG when your pass has been run.
1065 </p>
1067 <p>
1068 Some analyses chain to other analyses to do their job. For example, an <a
1071 cases where analyses chain, the <tt>addRequiredTransitive</tt> method should be
1072 used instead of the <tt>addRequired</tt> method. This informs the PassManager
1073 that the transitively required pass should be alive as long as the requiring
1074 pass is.
1075 </p>
1076 </div>
1078 <!-- _______________________________________________________________________ -->
1081 </div>
1084 <p>
1085 One of the jobs of the PassManager is to optimize how and when analyses are run.
1086 In particular, it attempts to avoid recomputing data unless it needs to. For
1087 this reason, passes are allowed to declare that they preserve (i.e., they don't
1088 invalidate) an existing analysis if it's available. For example, a simple
1089 constant folding pass would not modify the CFG, so it can't possibly affect the
1090 results of dominator analysis. By default, all passes are assumed to invalidate
1091 all others.
1092 </p>
1094 <p>
1095 The <tt>AnalysisUsage</tt> class provides several methods which are useful in
1096 certain circumstances that are related to <tt>addPreserved</tt>. In particular,
1097 the <tt>setPreservesAll</tt> method can be called to indicate that the pass does
1098 not modify the LLVM program at all (which is true for analyses), and the
1099 <tt>setPreservesCFG</tt> method can be used by transformations that change
1100 instructions in the program but do not modify the CFG or terminator instructions
1101 (note that this property is implicitly set for <a
1103 </p>
1105 <p>
1106 <tt>addPreserved</tt> is particularly useful for transformations like
1107 <tt>BreakCriticalEdges</tt>. This pass knows how to update a small set of loop
1108 and dominator related analyses if they exist, so it can preserve them, despite
1109 the fact that it hacks on the CFG.
1110 </p>
1111 </div>
1113 <!-- _______________________________________________________________________ -->
1116 </div>
1121 <i>// This is an example implementation from an analysis, which does not modify
1122 // the program at all, yet has a prerequisite.</i>
1123 <b>void</b> <a href="http://llvm.org/doxygen/classllvm_1_1PostDominanceFrontier.html">PostDominanceFrontier</a>::getAnalysisUsage(AnalysisUsage &AU) <b>const</b> {
1124 AU.setPreservesAll();
1125 AU.addRequired<<a href="http://llvm.org/doxygen/classllvm_1_1PostDominatorTree.html">PostDominatorTree</a>>();
1126 }
1127 </pre></div>
1129 <p>and:</p>
1132 <i>// This example modifies the program, but does not modify the CFG</i>
1133 <b>void</b> <a href="http://llvm.org/doxygen/structLICM.html">LICM</a>::getAnalysisUsage(AnalysisUsage &AU) <b>const</b> {
1134 AU.setPreservesCFG();
1135 AU.addRequired<<a href="http://llvm.org/doxygen/classllvm_1_1LoopInfo.html">LoopInfo</a>>();
1136 }
1137 </pre></div>
1139 </div>
1141 <!-- _______________________________________________________________________ -->
1144 <tt>getAnalysisIfAvailable<></tt> methods</a>
1145 </div>
1149 <p>The <tt>Pass::getAnalysis<></tt> method is automatically inherited by
1150 your class, providing you with access to the passes that you declared that you
1152 method. It takes a single template argument that specifies which pass class you
1153 want, and returns a reference to that pass. For example:</p>
1156 bool LICM::runOnFunction(Function &F) {
1157 LoopInfo &LI = getAnalysis<LoopInfo>();
1158 ...
1159 }
1160 </pre></div>
1162 <p>This method call returns a reference to the pass desired. You may get a
1163 runtime assertion failure if you attempt to get an analysis that you did not
1164 declare as required in your <a
1166 method can be called by your <tt>run*</tt> method implementation, or by any
1167 other local method invoked by your <tt>run*</tt> method.
1169 A module level pass can use function level analysis info using this interface.
1170 For example:</p>
1173 bool ModuleLevelPass::runOnModule(Module &M) {
1174 ...
1175 DominatorTree &DT = getAnalysis<DominatorTree>(Func);
1176 ...
1177 }
1178 </pre></div>
1180 <p>In above example, runOnFunction for DominatorTree is called by pass manager
1181 before returning a reference to the desired pass.</p>
1183 <p>
1184 If your pass is capable of updating analyses if they exist (e.g.,
1185 <tt>BreakCriticalEdges</tt>, as described above), you can use the
1186 <tt>getAnalysisIfAvailable</tt> method, which returns a pointer to the analysis
1187 if it is active. For example:</p>
1190 ...
1191 if (DominatorSet *DS = getAnalysisIfAvailable<DominatorSet>()) {
1192 <i>// A DominatorSet is active. This code will update it.</i>
1193 }
1194 ...
1195 </pre></div>
1197 </div>
1199 <!-- *********************************************************************** -->
1202 </div>
1203 <!-- *********************************************************************** -->
1207 <p>Now that we understand the basics of how passes are defined, how they are
1208 used, and how they are required from other passes, it's time to get a little bit
1209 fancier. All of the pass relationships that we have seen so far are very
1210 simple: one pass depends on one other specific pass to be run before it can run.
1211 For many applications, this is great, for others, more flexibility is
1212 required.</p>
1214 <p>In particular, some analyses are defined such that there is a single simple
1215 interface to the analysis results, but multiple ways of calculating them.
1216 Consider alias analysis for example. The most trivial alias analysis returns
1218 flow-sensitive, context-sensitive interprocedural analysis that can take a
1219 significant amount of time to execute (and obviously, there is a lot of room
1220 between these two extremes for other implementations). To cleanly support
1221 situations like this, the LLVM Pass Infrastructure supports the notion of
1222 Analysis Groups.</p>
1224 </div>
1226 <!-- _______________________________________________________________________ -->
1229 </div>
1233 <p>An Analysis Group is a single simple interface that may be implemented by
1234 multiple different passes. Analysis Groups can be given human readable names
1235 just like passes, but unlike passes, they need not derive from the <tt>Pass</tt>
1236 class. An analysis group may have one or more implementations, one of which is
1239 <p>Analysis groups are used by client passes just like other passes are: the
1240 <tt>AnalysisUsage::addRequired()</tt> and <tt>Pass::getAnalysis()</tt> methods.
1242 scans the available passes to see if any implementations of the analysis group
1243 are available. If none is available, the default implementation is created for
1245 between passes</a> still apply.</p>
1248 passes, all analysis group implementations must be registered, and must use the
1250 implementation pool. Also, a default implementation of the interface
1251 <b>must</b> be registered with <A
1254 <p>As a concrete example of an Analysis Group in action, consider the <a
1256 analysis group. The default implementation of the alias analysis interface (the
1257 <tt><a
1259 pass) just does a few simple checks that don't require significant analysis to
1260 compute (such as: two different globals can never alias each other, etc).
1261 Passes that use the <tt><a
1263 interface (for example the <tt><a
1265 not care which implementation of alias analysis is actually provided, they just
1266 use the designated interface.</p>
1268 <p>From the user's perspective, commands work just like normal. Issuing the
1269 command '<tt>opt -gcse ...</tt>' will cause the <tt>basicaa</tt> class to be
1270 instantiated and added to the pass sequence. Issuing the command '<tt>opt
1271 -somefancyaa -gcse ...</tt>' will cause the <tt>gcse</tt> pass to use the
1272 <tt>somefancyaa</tt> alias analysis (which doesn't actually exist, it's just a
1273 hypothetical example) instead.</p>
1275 </div>
1277 <!-- _______________________________________________________________________ -->
1280 </div>
1284 <p>The <tt>RegisterAnalysisGroup</tt> template is used to register the analysis
1285 group itself as well as add pass implementations to the analysis group. First,
1286 an analysis should be registered, with a human readable name provided for it.
1287 Unlike registration of passes, there is no command line argument to be specified
1291 <b>static</b> RegisterAnalysisGroup<<a href="http://llvm.org/doxygen/classllvm_1_1AliasAnalysis.html">AliasAnalysis</a>> A("<i>Alias Analysis</i>");
1292 </pre></div>
1294 <p>Once the analysis is registered, passes can declare that they are valid
1295 implementations of the interface by using the following code:</p>
1298 <b>namespace</b> {
1299 //<i> Analysis Group implementations <b>must</b> be registered normally...</i>
1300 RegisterPass<FancyAA>
1303 //<i> Declare that we implement the AliasAnalysis interface</i>
1304 RegisterAnalysisGroup<<a href="http://llvm.org/doxygen/classllvm_1_1AliasAnalysis.html">AliasAnalysis</a>> C(B);
1305 }
1306 </pre></div>
1308 <p>This just shows a class <tt>FancyAA</tt> that is registered normally, then
1311 analysis group. Every implementation of an analysis group should join using
1312 this template. A single pass may join multiple different analysis groups with
1313 no problem.</p>
1316 <b>namespace</b> {
1317 //<i> Analysis Group implementations <b>must</b> be registered normally...</i>
1318 RegisterPass<<a href="http://llvm.org/doxygen/structBasicAliasAnalysis.html">BasicAliasAnalysis</a>>
1321 //<i> Declare that we implement the AliasAnalysis interface</i>
1322 RegisterAnalysisGroup<<a href="http://llvm.org/doxygen/classllvm_1_1AliasAnalysis.html">AliasAnalysis</a>, <b>true</b>> E(D);
1323 }
1324 </pre></div>
1326 <p>Here we show how the default implementation is specified (using the extra
1327 argument to the <tt>RegisterAnalysisGroup</tt> template). There must be exactly
1328 one default implementation available at all times for an Analysis Group to be
1329 used. Only default implementation can derive from <tt>ImmutablePass</tt>.
1330 Here we declare that the
1331 <tt><a href="http://llvm.org/doxygen/structBasicAliasAnalysis.html">BasicAliasAnalysis</a></tt>
1332 pass is the default implementation for the interface.</p>
1334 </div>
1336 <!-- *********************************************************************** -->
1339 </div>
1340 <!-- *********************************************************************** -->
1343 <p>The <a
1345 class is designed to be an easy way to expose various success
1346 metrics from passes. These statistics are printed at the end of a
1347 run, when the -stats command line option is enabled on the command
1348 line. See the <a href="http://llvm.org/docs/ProgrammersManual.html#Statistic">Statistics section</a> in the Programmer's Manual for details.
1350 </div>
1353 <!-- *********************************************************************** -->
1356 </div>
1357 <!-- *********************************************************************** -->
1361 <p>The <a
1363 <a
1366 are set up correctly, and then schedules passes to run efficiently. All of the
1367 LLVM tools that run passes use the <tt>PassManager</tt> for execution of these
1368 passes.</p>
1370 <p>The <tt>PassManager</tt> does two main things to try to reduce the execution
1371 time of a series of passes:</p>
1373 <ol>
1374 <li><b>Share analysis results</b> - The PassManager attempts to avoid
1375 recomputing analysis results as much as possible. This means keeping track of
1376 which analyses are available already, which analyses get invalidated, and which
1377 analyses are needed to be run for a pass. An important part of work is that the
1378 <tt>PassManager</tt> tracks the exact lifetime of all analysis results, allowing
1380 results as soon as they are no longer needed.</li>
1382 <li><b>Pipeline the execution of passes on the program</b> - The
1383 <tt>PassManager</tt> attempts to get better cache and memory usage behavior out
1384 of a series of passes by pipelining the passes together. This means that, given
1387 the first function, then all of the <a
1389 etc... until the entire program has been run through the passes.
1391 <p>This improves the cache behavior of the compiler, because it is only touching
1392 the LLVM program representation for a single function at a time, instead of
1393 traversing the entire program. It reduces the memory consumption of compiler,
1394 because, for example, only one <a
1396 needs to be calculated at a time. This also makes it possible to implement
1397 some <a
1400 </ol>
1402 <p>The effectiveness of the <tt>PassManager</tt> is influenced directly by how
1403 much information it has about the behaviors of the passes it is scheduling. For
1406 Not implementing when it should be implemented will have the effect of not
1407 allowing any analysis results to live across the execution of your pass.</p>
1409 <p>The <tt>PassManager</tt> class exposes a <tt>--debug-pass</tt> command line
1410 options that is useful for debugging pass execution, seeing how things work, and
1411 diagnosing when you should be preserving more analyses than you currently are
1412 (To get information about all of the variants of the <tt>--debug-pass</tt>
1413 option, just type '<tt>opt --help-hidden</tt>').</p>
1415 <p>By using the <tt>--debug-pass=Structure</tt> option, for example, we can see
1417 Lets try it out with the <tt>gcse</tt> and <tt>licm</tt> passes:</p>
1420 $ opt -load ../../../Debug/lib/Hello.so -gcse -licm --debug-pass=Structure < hello.bc > /dev/null
1421 Module Pass Manager
1422 Function Pass Manager
1423 Dominator Set Construction
1424 Immediate Dominators Construction
1425 Global Common Subexpression Elimination
1426 -- Immediate Dominators Construction
1427 -- Global Common Subexpression Elimination
1428 Natural Loop Construction
1429 Loop Invariant Code Motion
1430 -- Natural Loop Construction
1431 -- Loop Invariant Code Motion
1432 Module Verifier
1433 -- Dominator Set Construction
1434 -- Module Verifier
1435 Bitcode Writer
1436 --Bitcode Writer
1437 </pre></div>
1439 <p>This output shows us when passes are constructed and when the analysis
1440 results are known to be dead (prefixed with '<tt>--</tt>'). Here we see that
1441 GCSE uses dominator and immediate dominator information to do its job. The LICM
1442 pass uses natural loop information, which uses dominator sets, but not immediate
1443 dominators. Because immediate dominators are no longer useful after the GCSE
1444 pass, it is immediately destroyed. The dominator sets are then reused to
1445 compute natural loop information, which is then used by the LICM pass.</p>
1447 <p>After the LICM pass, the module verifier runs (which is automatically added
1448 by the '<tt>opt</tt>' tool), which uses the dominator set to check that the
1449 resultant LLVM code is well formed. After it finishes, the dominator set
1450 information is destroyed, after being computed once, and shared by three
1451 passes.</p>
1454 World</a> pass in between the two passes:</p>
1457 $ opt -load ../../../Debug/lib/Hello.so -gcse -hello -licm --debug-pass=Structure < hello.bc > /dev/null
1458 Module Pass Manager
1459 Function Pass Manager
1460 Dominator Set Construction
1461 Immediate Dominators Construction
1462 Global Common Subexpression Elimination
1463 <b>-- Dominator Set Construction</b>
1464 -- Immediate Dominators Construction
1465 -- Global Common Subexpression Elimination
1466 <b> Hello World Pass
1467 -- Hello World Pass
1468 Dominator Set Construction</b>
1469 Natural Loop Construction
1470 Loop Invariant Code Motion
1471 -- Natural Loop Construction
1472 -- Loop Invariant Code Motion
1473 Module Verifier
1474 -- Dominator Set Construction
1475 -- Module Verifier
1476 Bitcode Writer
1477 --Bitcode Writer
1478 Hello: __main
1479 Hello: puts
1480 Hello: main
1481 </pre></div>
1484 Dominator Set pass, even though it doesn't modify the code at all! To fix this,
1485 we need to add the following <a
1489 <i>// We don't modify the program, so we preserve all analyses</i>
1490 <b>virtual void</b> getAnalysisUsage(AnalysisUsage &AU) <b>const</b> {
1491 AU.setPreservesAll();
1492 }
1493 </pre></div>
1495 <p>Now when we run our pass, we get this output:</p>
1498 $ opt -load ../../../Debug/lib/Hello.so -gcse -hello -licm --debug-pass=Structure < hello.bc > /dev/null
1499 Pass Arguments: -gcse -hello -licm
1500 Module Pass Manager
1501 Function Pass Manager
1502 Dominator Set Construction
1503 Immediate Dominators Construction
1504 Global Common Subexpression Elimination
1505 -- Immediate Dominators Construction
1506 -- Global Common Subexpression Elimination
1507 Hello World Pass
1508 -- Hello World Pass
1509 Natural Loop Construction
1510 Loop Invariant Code Motion
1511 -- Loop Invariant Code Motion
1512 -- Natural Loop Construction
1513 Module Verifier
1514 -- Dominator Set Construction
1515 -- Module Verifier
1516 Bitcode Writer
1517 --Bitcode Writer
1518 Hello: __main
1519 Hello: puts
1520 Hello: main
1521 </pre></div>
1523 <p>Which shows that we don't accidentally invalidate dominator information
1524 anymore, and therefore do not have to compute it twice.</p>
1526 </div>
1528 <!-- _______________________________________________________________________ -->
1531 </div>
1536 <b>virtual void</b> releaseMemory();
1537 </pre></div>
1539 <p>The <tt>PassManager</tt> automatically determines when to compute analysis
1540 results, and how long to keep them around for. Because the lifetime of the pass
1541 object itself is effectively the entire duration of the compilation process, we
1542 need some way to free analysis results when they are no longer useful. The
1543 <tt>releaseMemory</tt> virtual method is the way to do this.</p>
1545 <p>If you are writing an analysis or any other pass that retains a significant
1548 <tt>releaseMemory</tt> to, well, release the memory allocated to maintain this
1549 internal state. This method is called after the <tt>run*</tt> method for the
1550 class, before the next call of <tt>run*</tt> in your pass.</p>
1552 </div>
1554 <!-- *********************************************************************** -->
1557 </div>
1558 <!-- *********************************************************************** -->
1562 <p><i>Size matters</i> when constructing production quality tools using llvm,
1563 both for the purposes of distribution, and for regulating the resident code size
1564 when running on the target system. Therefore, it becomes desirable to
1565 selectively use some passes, while omitting others and maintain the flexibility
1566 to change configurations later on. You want to be able to do all this, and,
1567 provide feedback to the user. This is where pass registration comes into
1568 play.</p>
1570 <p>The fundamental mechanisms for pass registration are the
1571 <tt>MachinePassRegistry</tt> class and subclasses of
1572 <tt>MachinePassRegistryNode</tt>.</p>
1574 <p>An instance of <tt>MachinePassRegistry</tt> is used to maintain a list of
1575 <tt>MachinePassRegistryNode</tt> objects. This instance maintains the list and
1576 communicates additions and deletions to the command line interface.</p>
1578 <p>An instance of <tt>MachinePassRegistryNode</tt> subclass is used to maintain
1579 information provided about a particular pass. This information includes the
1580 command line name, the command help string and the address of the function used
1581 to create an instance of the pass. A global static constructor of one of these
1582 instances <i>registers</i> with a corresponding <tt>MachinePassRegistry</tt>,
1583 the static destructor <i>unregisters</i>. Thus a pass that is statically linked
1584 in the tool will be registered at start up. A dynamically loaded pass will
1585 register on load and unregister at unload.</p>
1587 </div>
1589 <!-- _______________________________________________________________________ -->
1592 </div>
1596 <p>There are predefined registries to track instruction scheduling
1597 (<tt>RegisterScheduler</tt>) and register allocation (<tt>RegisterRegAlloc</tt>)
1598 machine passes. Here we will describe how to <i>register</i> a register
1599 allocator machine pass.</p>
1601 <p>Implement your register allocator machine pass. In your register allocator
1602 .cpp file add the following include;</p>
1606 </pre></div>
1608 <p>Also in your register allocator .cpp file, define a creator function in the
1609 form; </p>
1612 FunctionPass *createMyRegisterAllocator() {
1613 return new MyRegisterAllocator();
1614 }
1615 </pre></div>
1617 <p>Note that the signature of this function should match the type of
1618 <tt>RegisterRegAlloc::FunctionPassCtor</tt>. In the same file add the
1624 createMyRegisterAllocator);
1625 </pre></div>
1627 <p>Note the two spaces prior to the help string produces a tidy result on the
1628 --help query.</p>
1631 $ llc --help
1632 ...
1633 -regalloc - Register allocator to use: (default = linearscan)
1634 =linearscan - linear scan register allocator
1635 =local - local register allocator
1636 =simple - simple register allocator
1637 =myregalloc - my register allocator help string
1638 ...
1639 </pre></div>
1641 <p>And that's it. The user is now free to use <tt>-regalloc=myregalloc</tt> as
1642 an option. Registering instruction schedulers is similar except use the
1643 <tt>RegisterScheduler</tt> class. Note that the
1644 <tt>RegisterScheduler::FunctionPassCtor</tt> is significantly different from
1645 <tt>RegisterRegAlloc::FunctionPassCtor</tt>.</p>
1647 <p>To force the load/linking of your register allocator into the llc/lli tools,
1649 call line to <tt>llvm/Codegen/LinkAllCodegenComponents.h</tt>.</p>
1651 </div>
1654 <!-- _______________________________________________________________________ -->
1657 </div>
1661 <p>The easiest way to get started is to clone one of the existing registries; we
1662 recommend <tt>llvm/CodeGen/RegAllocRegistry.h</tt>. The key things to modify
1663 are the class name and the <tt>FunctionPassCtor</tt> type.</p>
1665 <p>Then you need to declare the registry. Example: if your pass registry is
1666 <tt>RegisterMyPasses</tt> then define;</p>
1669 MachinePassRegistry RegisterMyPasses::Registry;
1670 </pre></div>
1672 <p>And finally, declare the command line option for your passes. Example:</p>
1675 cl::opt<RegisterMyPasses::FunctionPassCtor, false,
1676 RegisterPassParser<RegisterMyPasses> >
1678 cl::init(&createDefaultMyPass),
1680 </pre></div>
1683 creator.</p>
1685 </div>
1687 <!-- *********************************************************************** -->
1690 </div>
1691 <!-- *********************************************************************** -->
1695 <p>Unfortunately, using GDB with dynamically loaded passes is not as easy as it
1696 should be. First of all, you can't set a breakpoint in a shared object that has
1697 not been loaded yet, and second of all there are problems with inlined functions
1698 in shared objects. Here are some suggestions to debugging your pass with
1699 GDB.</p>
1701 <p>For sake of discussion, I'm going to assume that you are debugging a
1702 transformation invoked by <tt>opt</tt>, although nothing described here depends
1703 on that.</p>
1705 </div>
1707 <!-- _______________________________________________________________________ -->
1710 </div>
1714 <p>First thing you do is start <tt>gdb</tt> on the <tt>opt</tt> process:</p>
1717 $ <b>gdb opt</b>
1718 GNU gdb 5.0
1719 Copyright 2000 Free Software Foundation, Inc.
1720 GDB is free software, covered by the GNU General Public License, and you are
1721 welcome to change it and/or distribute copies of it under certain conditions.
1725 (gdb)
1726 </pre></div>
1728 <p>Note that <tt>opt</tt> has a lot of debugging information in it, so it takes
1729 time to load. Be patient. Since we cannot set a breakpoint in our pass yet
1730 (the shared object isn't loaded until runtime), we must execute the process, and
1731 have it stop before it invokes our pass, but after it has loaded the shared
1732 object. The most foolproof way of doing this is to set a breakpoint in
1733 <tt>PassManager::run</tt> and then run the process with the arguments you
1734 want:</p>
1737 (gdb) <b>break llvm::PassManager::run</b>
1738 Breakpoint 1 at 0x2413bc: file Pass.cpp, line 70.
1739 (gdb) <b>run test.bc -load $(LLVMTOP)/llvm/Debug/lib/[libname].so -[passoption]</b>
1740 Starting program: opt test.bc -load $(LLVMTOP)/llvm/Debug/lib/[libname].so -[passoption]
1741 Breakpoint 1, PassManager::run (this=0xffbef174, M=@0x70b298) at Pass.cpp:70
1742 70 bool PassManager::run(Module &M) { return PM->run(M); }
1743 (gdb)
1744 </pre></div>
1746 <p>Once the <tt>opt</tt> stops in the <tt>PassManager::run</tt> method you are
1747 now free to set breakpoints in your pass so that you can trace through execution
1748 or do other standard debugging stuff.</p>
1750 </div>
1752 <!-- _______________________________________________________________________ -->
1755 </div>
1759 <p>Once you have the basics down, there are a couple of problems that GDB has,
1760 some with solutions, some without.</p>
1762 <ul>
1763 <li>Inline functions have bogus stack information. In general, GDB does a
1764 pretty good job getting stack traces and stepping through inline functions.
1765 When a pass is dynamically loaded however, it somehow completely loses this
1766 capability. The only solution I know of is to de-inline a function (move it
1767 from the body of a class to a .cpp file).</li>
1769 <li>Restarting the program breaks breakpoints. After following the information
1770 above, you have succeeded in getting some breakpoints planted in your pass. Nex
1771 thing you know, you restart the program (i.e., you type '<tt>run</tt>' again),
1772 and you start getting errors about breakpoints being unsettable. The only way I
1774 already set in your pass, run the program, and re-set the breakpoints once
1775 execution stops in <tt>PassManager::run</tt>.</li>
1777 </ul>
1779 <p>Hopefully these tips will help with common case debugging situations. If
1780 you'd like to contribute some tips of your own, just contact <a
1783 </div>
1785 <!-- *********************************************************************** -->
1788 </div>
1789 <!-- *********************************************************************** -->
1793 <p>Although the LLVM Pass Infrastructure is very capable as it stands, and does
1794 some nifty stuff, there are things we'd like to add in the future. Here is
1795 where we are going:</p>
1797 </div>
1799 <!-- _______________________________________________________________________ -->
1802 </div>
1806 <p>Multiple CPU machines are becoming more common and compilation can never be
1807 fast enough: obviously we should allow for a multithreaded compiler. Because of
1808 the semantics defined for passes above (specifically they cannot maintain state
1809 across invocations of their <tt>run*</tt> methods), a nice clean way to
1810 implement a multithreaded compiler would be for the <tt>PassManager</tt> class
1811 to create multiple instances of each pass object, and allow the separate
1812 instances to be hacking on different parts of the program at the same time.</p>
1814 <p>This implementation would prevent each of the passes from having to implement
1815 multithreaded constructs, requiring only the LLVM core to have locking in a few
1816 places (for global resources). Although this is a simple extension, we simply
1817 haven't had time (or multiprocessor machines, thus a reason) to implement this.
1818 Despite that, we have kept the LLVM passes SMP ready, and you should too.</p>
1820 </div>
1822 <!-- *********************************************************************** -->
1823 <hr>
1824 <address>
1832 Last modified: $Date$
1833 </address>
1835 </body>
1836 </html>