| blob | 3cef2c934482485586b113e27b6416bc2f288734 |
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 hierarchy
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 # Include the makefile implementation stuff
193 include $(LEVEL)/Makefile.common
194 </pre></div>
197 directory are to be compiled and linked together into a
200 If your operating system uses a suffix other than .so (such as windows or
201 Mac OS/X), the appropriate extension will be used.</p>
203 <p>Now that we have the build scripts set up, we just need to write the code for
204 the pass itself.</p>
206 </div>
208 <!-- ======================================================================= -->
211 </div>
215 <p>Now that we have a way to compile our new pass, we just have to write it.
216 Start out with:</p>
221 <b>#include</b> "<a href="http://llvm.org/doxygen/raw__ostream_8h.html">llvm/Support/raw_ostream.h</a>"
222 </pre></div>
226 we are operating on <tt><a
228 and we will be doing some printing.</p>
233 </pre></div>
234 <p>... which is required because the functions from the include files
235 live in the llvm namespace.
236 </p>
242 </pre></div>
244 <p>... which starts out an anonymous namespace. Anonymous namespaces are to C++
245 what the "<tt>static</tt>" keyword is to C (at global scope). It makes the
246 things declared inside of the anonymous namespace only visible to the current
247 file. If you're not familiar with them, consult a decent C++ book for more
248 information.</p>
254 </pre></div><p>
258 The different builtin pass subclasses are described in detail <a
261 time.</p>
264 static char ID;
265 Hello() : FunctionPass(&ID) {}
266 </pre></div><p>
268 <p> This declares pass identifier used by LLVM to identify pass. This allows LLVM to
269 avoid using expensive C++ runtime information.</p>
275 }
277 </pre></div>
280 which overloads an abstract virtual method inherited from <a
282 to do our thing, so we just print out our message with the name of each
283 function.</p>
286 char Hello::ID = 0;
287 </pre></div>
289 <p> We initialize pass ID here. LLVM uses ID's address to identify pass so
290 initialization value is not important.</p>
294 false /* Only looks at CFG */,
295 false /* Analysis Pass */);
297 </pre></div>
300 giving it a command line
302 Last two arguments describe its behavior.
303 If a pass walks CFG without modifying it then third argument is set to true.
304 If a pass is an analysis pass, for example dominator tree pass, then true
305 is supplied as fourth argument. </p>
312 <b>#include</b> "<a href="http://llvm.org/doxygen/raw__ostream_8h.html">llvm/Support/raw_ostream.h</a>"
319 static char ID;
320 Hello() : FunctionPass(&ID) {}
325 }
326 };
328 char Hello::ID = 0;
330 }
332 </pre></div>
335 command in the local directory and you should get a new
336 "<tt>Debug+Asserts/lib/Hello.so</tt> file. Note that everything in this file is
337 contained in an anonymous namespace: this reflects the fact that passes are self
338 contained units that do not need external interfaces (although they can have
339 them) to be useful.</p>
341 </div>
343 <!-- ======================================================================= -->
346 </div>
350 <p>Now that you have a brand new shiny shared object file, we can use the
351 <tt>opt</tt> command to run an LLVM program through your pass. Because you
352 registered your pass with the <tt>INITIALIZE_PASS</tt> macro, you will be able to
353 use the <tt>opt</tt> tool to access it, once loaded.</p>
355 <p>To test it, follow the example at the end of the <a
357 LLVM. We can now run the bitcode file (<tt>hello.bc</tt>) for the program
358 through our transformation like this (or course, any bitcode file will
359 work):</p>
362 $ opt -load ../../../Debug+Asserts/lib/Hello.so -hello < hello.bc > /dev/null
363 Hello: __main
364 Hello: puts
365 Hello: main
366 </pre></div>
368 <p>The '<tt>-load</tt>' option specifies that '<tt>opt</tt>' should load your
369 pass as a shared object, which makes '<tt>-hello</tt>' a valid command line
371 pass</a>). Because the hello pass does not modify the program in any
372 interesting way, we just throw away the result of <tt>opt</tt> (sending it to
373 <tt>/dev/null</tt>).</p>
375 <p>To see what happened to the other string you registered, try running
376 <tt>opt</tt> with the <tt>-help</tt> option:</p>
379 $ opt -load ../../../Debug+Asserts/lib/Hello.so -help
380 OVERVIEW: llvm .bc -> .bc modular optimizer
382 USAGE: opt [options] <input bitcode>
384 OPTIONS:
385 Optimizations available:
386 ...
387 -funcresolve - Resolve Functions
388 -gcse - Global Common Subexpression Elimination
389 -globaldce - Dead Global Elimination
390 <b>-hello - Hello World Pass</b>
391 -indvars - Canonicalize Induction Variables
392 -inline - Function Integration/Inlining
393 -instcombine - Combine redundant instructions
394 ...
395 </pre></div>
397 <p>The pass name get added as the information string for your pass, giving some
398 documentation to users of <tt>opt</tt>. Now that you have a working pass, you
399 would go ahead and make it do the cool transformations you want. Once you get
400 it all working and tested, it may become useful to find out how fast your pass
402 line option (<tt>--time-passes</tt>) that allows you to get information about
403 the execution time of your pass along with the other passes you queue up. For
404 example:</p>
407 $ opt -load ../../../Debug+Asserts/lib/Hello.so -hello -time-passes < hello.bc > /dev/null
408 Hello: __main
409 Hello: puts
410 Hello: main
411 ===============================================================================
412 ... Pass execution timing report ...
413 ===============================================================================
414 Total Execution Time: 0.02 seconds (0.0479059 wall clock)
416 ---User Time--- --System Time-- --User+System-- ---Wall Time--- --- Pass Name ---
417 0.0100 (100.0%) 0.0000 ( 0.0%) 0.0100 ( 50.0%) 0.0402 ( 84.0%) Bitcode Writer
418 0.0000 ( 0.0%) 0.0100 (100.0%) 0.0100 ( 50.0%) 0.0031 ( 6.4%) Dominator Set Construction
419 0.0000 ( 0.0%) 0.0000 ( 0.0%) 0.0000 ( 0.0%) 0.0013 ( 2.7%) Module Verifier
420 <b> 0.0000 ( 0.0%) 0.0000 ( 0.0%) 0.0000 ( 0.0%) 0.0033 ( 6.9%) Hello World Pass</b>
421 0.0100 (100.0%) 0.0100 (100.0%) 0.0200 (100.0%) 0.0479 (100.0%) TOTAL
422 </pre></div>
424 <p>As you can see, our implementation above is pretty fast :). The additional
425 passes listed are automatically inserted by the '<tt>opt</tt>' tool to verify
426 that the LLVM emitted by your pass is still valid and well formed LLVM, which
427 hasn't been broken somehow.</p>
429 <p>Now that you have seen the basics of the mechanics behind passes, we can talk
430 about some more details of how they work and how to use them.</p>
432 </div>
434 <!-- *********************************************************************** -->
437 </div>
438 <!-- *********************************************************************** -->
442 <p>One of the first things that you should do when designing a new pass is to
443 decide what class you should subclass for your pass. The <a
446 did not discuss why or when this should occur. Here we talk about the classes
447 available, from the most general to the most specific.</p>
449 <p>When choosing a superclass for your Pass, you should choose the <b>most
450 specific</b> class possible, while still being able to meet the requirements
451 listed. This gives the LLVM Pass Infrastructure information necessary to
452 optimize how passes are run, so that the resultant compiler isn't unnecessarily
453 slow.</p>
455 </div>
457 <!-- ======================================================================= -->
460 </div>
466 class. This pass type is used for passes that do not have to be run, do not
467 change state, and never need to be updated. This is not a normal type of
468 transformation or analysis, but can provide information about the current
469 compiler configuration.</p>
471 <p>Although this pass class is very infrequently used, it is important for
472 providing information about the current target machine being compiled for, and
473 other static information that can affect the various transformations.</p>
475 <p><tt>ImmutablePass</tt>es never invalidate other transformations, are never
478 </div>
480 <!-- ======================================================================= -->
483 </div>
489 class is the most general of all superclasses that you can use. Deriving from
490 <tt>ModulePass</tt> indicates that your pass uses the entire program as a unit,
491 referring to function bodies in no predictable order, or adding and removing
492 functions. Because nothing is known about the behavior of <tt>ModulePass</tt>
493 subclasses, no optimization can be done for their execution.</p>
495 <p>A module pass can use function level passes (e.g. dominators) using
496 the getAnalysis interface
497 <tt>getAnalysis<DominatorTree>(llvm::Function *)</tt> to provide the
498 function to retrieve analysis result for, if the function pass does not require
499 any module or immutable passes. Note that this can only be done for functions for which the
500 analysis ran, e.g. in the case of dominators you should only ask for the
501 DominatorTree for function definitions, not declarations.</p>
503 <p>To write a correct <tt>ModulePass</tt> subclass, derive from
504 <tt>ModulePass</tt> and overload the <tt>runOnModule</tt> method with the
505 following signature:</p>
507 </div>
509 <!-- _______________________________________________________________________ -->
512 </div>
517 <b>virtual bool</b> runOnModule(Module &M) = 0;
518 </pre></div>
520 <p>The <tt>runOnModule</tt> method performs the interesting work of the pass.
521 It should return true if the module was modified by the transformation and
522 false otherwise.</p>
524 </div>
526 <!-- ======================================================================= -->
529 </div>
535 is used by passes that need to traverse the program bottom-up on the call graph
536 (callees before callers). Deriving from CallGraphSCCPass provides some
537 mechanics for building and traversing the CallGraph, but also allows the system
538 to optimize execution of CallGraphSCCPass's. If your pass meets the
539 requirements outlined below, and doesn't meet the requirements of a <tt><a
542 <tt>CallGraphSCCPass</tt>.</p>
544 <p><b>TODO</b>: explain briefly what SCC, Tarjan's algo, and B-U mean.</p>
546 <p>To be explicit, <tt>CallGraphSCCPass</tt> subclasses are:</p>
548 <ol>
550 <li>... <em>not allowed</em> to modify any <tt>Function</tt>s that are not in
551 the current SCC.</li>
553 <li>... <em>not allowed</em> to inspect any Function's other than those in the
554 current SCC and the direct callees of the SCC.</li>
556 <li>... <em>required</em> to preserve the current CallGraph object, updating it
557 to reflect any changes made to the program.</li>
559 <li>... <em>not allowed</em> to add or remove SCC's from the current Module,
560 though they may change the contents of an SCC.</li>
562 <li>... <em>allowed</em> to add or remove global variables from the current
563 Module.</li>
565 <li>... <em>allowed</em> to maintain state across invocations of
567 </ol>
569 <p>Implementing a <tt>CallGraphSCCPass</tt> is slightly tricky in some cases
570 because it has to handle SCCs with more than one node in it. All of the virtual
571 methods described below should return true if they modified the program, or
572 false if they didn't.</p>
574 </div>
576 <!-- _______________________________________________________________________ -->
579 method</a>
580 </div>
585 <b>virtual bool</b> doInitialization(CallGraph &CG);
586 </pre></div>
588 <p>The <tt>doIninitialize</tt> method is allowed to do most of the things that
589 <tt>CallGraphSCCPass</tt>'s are not allowed to do. They can add and remove
590 functions, get pointers to functions, etc. The <tt>doInitialization</tt> method
591 is designed to do simple initialization type of stuff that does not depend on
592 the SCCs being processed. The <tt>doInitialization</tt> method call is not
593 scheduled to overlap with any other pass executions (thus it should be very
594 fast).</p>
596 </div>
598 <!-- _______________________________________________________________________ -->
601 </div>
606 <b>virtual bool</b> runOnSCC(CallGraphSCC &SCC) = 0;
607 </pre></div>
609 <p>The <tt>runOnSCC</tt> method performs the interesting work of the pass, and
610 should return true if the module was modified by the transformation, false
611 otherwise.</p>
613 </div>
615 <!-- _______________________________________________________________________ -->
618 &)</tt> method</a>
619 </div>
624 <b>virtual bool</b> doFinalization(CallGraph &CG);
625 </pre></div>
627 <p>The <tt>doFinalization</tt> method is an infrequently used method that is
628 called when the pass framework has finished calling <a
630 program being compiled.</p>
632 </div>
634 <!-- ======================================================================= -->
637 </div>
641 <p>In contrast to <tt>ModulePass</tt> subclasses, <tt><a
643 subclasses do have a predictable, local behavior that can be expected by the
644 system. All <tt>FunctionPass</tt> execute on each function in the program
645 independent of all of the other functions in the program.
646 <tt>FunctionPass</tt>'s do not require that they are executed in a particular
647 order, and <tt>FunctionPass</tt>'s do not modify external functions.</p>
649 <p>To be explicit, <tt>FunctionPass</tt> subclasses are not allowed to:</p>
651 <ol>
652 <li>Modify a Function other than the one currently being processed.</li>
653 <li>Add or remove Function's from the current Module.</li>
654 <li>Add or remove global variables from the current Module.</li>
655 <li>Maintain state across invocations of
657 </ol>
659 <p>Implementing a <tt>FunctionPass</tt> is usually straightforward (See the <a
661 may overload three virtual methods to do their work. All of these methods
662 should return true if they modified the program, or false if they didn't.</p>
664 </div>
666 <!-- _______________________________________________________________________ -->
669 method</a>
670 </div>
675 <b>virtual bool</b> doInitialization(Module &M);
676 </pre></div>
678 <p>The <tt>doIninitialize</tt> method is allowed to do most of the things that
679 <tt>FunctionPass</tt>'s are not allowed to do. They can add and remove
680 functions, get pointers to functions, etc. The <tt>doInitialization</tt> method
681 is designed to do simple initialization type of stuff that does not depend on
682 the functions being processed. The <tt>doInitialization</tt> method call is not
683 scheduled to overlap with any other pass executions (thus it should be very
684 fast).</p>
686 <p>A good example of how this method should be used is the <a
688 pass. This pass converts <tt>malloc</tt> and <tt>free</tt> instructions into
689 platform dependent <tt>malloc()</tt> and <tt>free()</tt> function calls. It
690 uses the <tt>doInitialization</tt> method to get a reference to the malloc and
691 free functions that it needs, adding prototypes to the module if necessary.</p>
693 </div>
695 <!-- _______________________________________________________________________ -->
698 </div>
703 <b>virtual bool</b> runOnFunction(Function &F) = 0;
704 </pre></div><p>
706 <p>The <tt>runOnFunction</tt> method must be implemented by your subclass to do
707 the transformation or analysis work of your pass. As usual, a true value should
708 be returned if the function is modified.</p>
710 </div>
712 <!-- _______________________________________________________________________ -->
715 &)</tt> method</a>
716 </div>
721 <b>virtual bool</b> doFinalization(Module &M);
722 </pre></div>
724 <p>The <tt>doFinalization</tt> method is an infrequently used method that is
725 called when the pass framework has finished calling <a
727 program being compiled.</p>
729 </div>
731 <!-- ======================================================================= -->
734 </div>
738 <p> All <tt>LoopPass</tt> execute on each loop in the function independent of
739 all of the other loops in the function. <tt>LoopPass</tt> processes loops in
740 loop nest order such that outer most loop is processed last. </p>
742 <p> <tt>LoopPass</tt> subclasses are allowed to update loop nest using
743 <tt>LPPassManager</tt> interface. Implementing a loop pass is usually
744 straightforward. <tt>Looppass</tt>'s may overload three virtual methods to
745 do their work. All these methods should return true if they modified the
746 program, or false if they didn't. </p>
747 </div>
749 <!-- _______________________________________________________________________ -->
752 LPPassManager &)</tt>
753 method</a>
754 </div>
759 <b>virtual bool</b> doInitialization(Loop *, LPPassManager &LPM);
760 </pre></div>
762 <p>The <tt>doInitialization</tt> method is designed to do simple initialization
763 type of stuff that does not depend on the functions being processed. The
764 <tt>doInitialization</tt> method call is not scheduled to overlap with any
765 other pass executions (thus it should be very fast). LPPassManager
766 interface should be used to access Function or Module level analysis
767 information.</p>
769 </div>
772 <!-- _______________________________________________________________________ -->
775 </div>
780 <b>virtual bool</b> runOnLoop(Loop *, LPPassManager &LPM) = 0;
781 </pre></div><p>
783 <p>The <tt>runOnLoop</tt> method must be implemented by your subclass to do
784 the transformation or analysis work of your pass. As usual, a true value should
785 be returned if the function is modified. <tt>LPPassManager</tt> interface
786 should be used to update loop nest.</p>
788 </div>
790 <!-- _______________________________________________________________________ -->
793 </div>
798 <b>virtual bool</b> doFinalization();
799 </pre></div>
801 <p>The <tt>doFinalization</tt> method is an infrequently used method that is
802 called when the pass framework has finished calling <a
804 program being compiled. </p>
806 </div>
810 <!-- ======================================================================= -->
813 </div>
817 <p><tt>BasicBlockPass</tt>'s are just like <a
819 their scope of inspection and modification to a single basic block at a time.
820 As such, they are <b>not</b> allowed to do any of the following:</p>
822 <ol>
823 <li>Modify or inspect any basic blocks outside of the current one</li>
824 <li>Maintain state across invocations of
826 <li>Modify the control flow graph (by altering terminator instructions)</li>
827 <li>Any of the things forbidden for
829 </ol>
832 optimizations. They may override the same <a
835 href="#FunctionPass"><tt>FunctionPass</tt></a>'s have, but also have the following virtual methods that may also be implemented:</p>
837 </div>
839 <!-- _______________________________________________________________________ -->
842 &)</tt> method</a>
843 </div>
848 <b>virtual bool</b> doInitialization(Function &F);
849 </pre></div>
851 <p>The <tt>doIninitialize</tt> method is allowed to do most of the things that
852 <tt>BasicBlockPass</tt>'s are not allowed to do, but that
853 <tt>FunctionPass</tt>'s can. The <tt>doInitialization</tt> method is designed
854 to do simple initialization that does not depend on the
855 BasicBlocks being processed. The <tt>doInitialization</tt> method call is not
856 scheduled to overlap with any other pass executions (thus it should be very
857 fast).</p>
859 </div>
861 <!-- _______________________________________________________________________ -->
864 </div>
869 <b>virtual bool</b> runOnBasicBlock(BasicBlock &BB) = 0;
870 </pre></div>
872 <p>Override this function to do the work of the <tt>BasicBlockPass</tt>. This
873 function is not allowed to inspect or modify basic blocks other than the
874 parameter, and are not allowed to modify the CFG. A true value must be returned
875 if the basic block is modified.</p>
877 </div>
879 <!-- _______________________________________________________________________ -->
882 method</a>
883 </div>
888 <b>virtual bool</b> doFinalization(Function &F);
889 </pre></div>
891 <p>The <tt>doFinalization</tt> method is an infrequently used method that is
892 called when the pass framework has finished calling <a
894 program being compiled. This can be used to perform per-function
895 finalization.</p>
897 </div>
899 <!-- ======================================================================= -->
902 </div>
906 <p>A <tt>MachineFunctionPass</tt> is a part of the LLVM code generator that
907 executes on the machine-dependent representation of each LLVM function in the
908 program.</p>
910 <p>Code generator passes are registered and initialized specially by
911 <tt>TargetMachine::addPassesToEmitFile</tt> and similar routines, so they
912 cannot generally be run from the <tt>opt</tt> or <tt>bugpoint</tt>
913 commands.</p>
915 <p>A <tt>MachineFunctionPass</tt> is also a <tt>FunctionPass</tt>, so all
916 the restrictions that apply to a <tt>FunctionPass</tt> also apply to it.
917 <tt>MachineFunctionPass</tt>es also have additional restrictions. In particular,
918 <tt>MachineFunctionPass</tt>es are not allowed to do any of the following:</p>
920 <ol>
921 <li>Modify or create any LLVM IR Instructions, BasicBlocks, Arguments,
922 Functions, GlobalVariables, GlobalAliases, or Modules.</li>
923 <li>Modify a MachineFunction other than the one currently being processed.</li>
924 <li>Maintain state across invocations of <a
926 data)</li>
927 </ol>
929 </div>
931 <!-- _______________________________________________________________________ -->
934 &MF)</tt> method</a>
935 </div>
940 <b>virtual bool</b> runOnMachineFunction(MachineFunction &MF) = 0;
941 </pre></div>
943 <p><tt>runOnMachineFunction</tt> can be considered the main entry point of a
944 <tt>MachineFunctionPass</tt>; that is, you should override this method to do the
945 work of your <tt>MachineFunctionPass</tt>.</p>
947 <p>The <tt>runOnMachineFunction</tt> method is called on every
948 <tt>MachineFunction</tt> in a <tt>Module</tt>, so that the
949 <tt>MachineFunctionPass</tt> may perform optimizations on the machine-dependent
950 representation of the function. If you want to get at the LLVM <tt>Function</tt>
951 for the <tt>MachineFunction</tt> you're working on, use
952 <tt>MachineFunction</tt>'s <tt>getFunction()</tt> accessor method -- but
953 remember, you may not modify the LLVM <tt>Function</tt> or its contents from a
954 <tt>MachineFunctionPass</tt>.</p>
956 </div>
958 <!-- *********************************************************************** -->
961 </div>
962 <!-- *********************************************************************** -->
967 pass registration works, and discussed some of the reasons that it is used and
968 what it does. Here we discuss how and why passes are registered.</p>
970 <p>As we saw above, passes are registered with the <b><tt>INITIALIZE_PASS</tt></b>
971 macro. The first parameter is the name of the pass that is to be used on
972 the command line to specify that the pass should be added to a program (for
973 example, with <tt>opt</tt> or <tt>bugpoint</tt>). The second argument is the
974 name of the pass, which is to be used for the <tt>-help</tt> output of
975 programs, as
976 well as for debug output generated by the <tt>--debug-pass</tt> option.</p>
978 <p>If you want your pass to be easily dumpable, you should
979 implement the virtual <tt>print</tt> method:</p>
981 </div>
983 <!-- _______________________________________________________________________ -->
986 </div>
991 <b>virtual void</b> print(std::ostream &O, <b>const</b> Module *M) <b>const</b>;
992 </pre></div>
995 a human readable version of the analysis results. This is useful for debugging
996 an analysis itself, as well as for other people to figure out how an analysis
997 works. Use the <tt>opt -analyze</tt> argument to invoke this method.</p>
999 <p>The <tt>llvm::OStream</tt> parameter specifies the stream to write the results on,
1000 and the <tt>Module</tt> parameter gives a pointer to the top level module of the
1001 program that has been analyzed. Note however that this pointer may be null in
1002 certain circumstances (such as calling the <tt>Pass::dump()</tt> from a
1003 debugger), so it should only be used to enhance debug output, it should not be
1004 depended on.</p>
1006 </div>
1008 <!-- *********************************************************************** -->
1011 </div>
1012 <!-- *********************************************************************** -->
1016 <p>One of the main responsibilities of the <tt>PassManager</tt> is to make sure
1017 that passes interact with each other correctly. Because <tt>PassManager</tt>
1019 know how the passes interact with each other and what dependencies exist between
1020 the various passes. To track this, each pass can declare the set of passes that
1021 are required to be executed before the current pass, and the passes which are
1022 invalidated by the current pass.</p>
1024 <p>Typically this functionality is used to require that analysis results are
1025 computed before your pass is run. Running arbitrary transformation passes can
1026 invalidate the computed analysis results, which is what the invalidation set
1027 specifies. If a pass does not implement the <tt><a
1029 having any prerequisite passes, and invalidating <b>all</b> other passes.</p>
1031 </div>
1033 <!-- _______________________________________________________________________ -->
1036 </div>
1041 <b>virtual void</b> getAnalysisUsage(AnalysisUsage &Info) <b>const</b>;
1042 </pre></div>
1044 <p>By implementing the <tt>getAnalysisUsage</tt> method, the required and
1045 invalidated sets may be specified for your transformation. The implementation
1046 should fill in the <tt><a
1048 object with information about which passes are required and not invalidated. To
1049 do this, a pass may call any of the following methods on the AnalysisUsage
1050 object:</p>
1051 </div>
1053 <!-- _______________________________________________________________________ -->
1055 <a name="AU::addRequired">The <tt>AnalysisUsage::addRequired<></tt> and <tt>AnalysisUsage::addRequiredTransitive<></tt> methods</a>
1056 </div>
1059 <p>
1060 If your pass requires a previous pass to be executed (an analysis for example),
1061 it can use one of these methods to arrange for it to be run before your pass.
1062 LLVM has many different types of analyses and passes that can be required,
1063 spanning the range from <tt>DominatorSet</tt> to <tt>BreakCriticalEdges</tt>.
1064 Requiring <tt>BreakCriticalEdges</tt>, for example, guarantees that there will
1065 be no critical edges in the CFG when your pass has been run.
1066 </p>
1068 <p>
1069 Some analyses chain to other analyses to do their job. For example, an <a
1072 cases where analyses chain, the <tt>addRequiredTransitive</tt> method should be
1073 used instead of the <tt>addRequired</tt> method. This informs the PassManager
1074 that the transitively required pass should be alive as long as the requiring
1075 pass is.
1076 </p>
1077 </div>
1079 <!-- _______________________________________________________________________ -->
1082 </div>
1085 <p>
1086 One of the jobs of the PassManager is to optimize how and when analyses are run.
1087 In particular, it attempts to avoid recomputing data unless it needs to. For
1088 this reason, passes are allowed to declare that they preserve (i.e., they don't
1089 invalidate) an existing analysis if it's available. For example, a simple
1090 constant folding pass would not modify the CFG, so it can't possibly affect the
1091 results of dominator analysis. By default, all passes are assumed to invalidate
1092 all others.
1093 </p>
1095 <p>
1096 The <tt>AnalysisUsage</tt> class provides several methods which are useful in
1097 certain circumstances that are related to <tt>addPreserved</tt>. In particular,
1098 the <tt>setPreservesAll</tt> method can be called to indicate that the pass does
1099 not modify the LLVM program at all (which is true for analyses), and the
1100 <tt>setPreservesCFG</tt> method can be used by transformations that change
1101 instructions in the program but do not modify the CFG or terminator instructions
1102 (note that this property is implicitly set for <a
1104 </p>
1106 <p>
1107 <tt>addPreserved</tt> is particularly useful for transformations like
1108 <tt>BreakCriticalEdges</tt>. This pass knows how to update a small set of loop
1109 and dominator related analyses if they exist, so it can preserve them, despite
1110 the fact that it hacks on the CFG.
1111 </p>
1112 </div>
1114 <!-- _______________________________________________________________________ -->
1117 </div>
1122 <i>// This is an example implementation from an analysis, which does not modify
1123 // the program at all, yet has a prerequisite.</i>
1124 <b>void</b> <a href="http://llvm.org/doxygen/classllvm_1_1PostDominanceFrontier.html">PostDominanceFrontier</a>::getAnalysisUsage(AnalysisUsage &AU) <b>const</b> {
1125 AU.setPreservesAll();
1126 AU.addRequired<<a href="http://llvm.org/doxygen/classllvm_1_1PostDominatorTree.html">PostDominatorTree</a>>();
1127 }
1128 </pre></div>
1130 <p>and:</p>
1133 <i>// This example modifies the program, but does not modify the CFG</i>
1134 <b>void</b> <a href="http://llvm.org/doxygen/structLICM.html">LICM</a>::getAnalysisUsage(AnalysisUsage &AU) <b>const</b> {
1135 AU.setPreservesCFG();
1136 AU.addRequired<<a href="http://llvm.org/doxygen/classllvm_1_1LoopInfo.html">LoopInfo</a>>();
1137 }
1138 </pre></div>
1140 </div>
1142 <!-- _______________________________________________________________________ -->
1145 <tt>getAnalysisIfAvailable<></tt> methods</a>
1146 </div>
1150 <p>The <tt>Pass::getAnalysis<></tt> method is automatically inherited by
1151 your class, providing you with access to the passes that you declared that you
1153 method. It takes a single template argument that specifies which pass class you
1154 want, and returns a reference to that pass. For example:</p>
1157 bool LICM::runOnFunction(Function &F) {
1158 LoopInfo &LI = getAnalysis<LoopInfo>();
1159 ...
1160 }
1161 </pre></div>
1163 <p>This method call returns a reference to the pass desired. You may get a
1164 runtime assertion failure if you attempt to get an analysis that you did not
1165 declare as required in your <a
1167 method can be called by your <tt>run*</tt> method implementation, or by any
1168 other local method invoked by your <tt>run*</tt> method.
1170 A module level pass can use function level analysis info using this interface.
1171 For example:</p>
1174 bool ModuleLevelPass::runOnModule(Module &M) {
1175 ...
1176 DominatorTree &DT = getAnalysis<DominatorTree>(Func);
1177 ...
1178 }
1179 </pre></div>
1181 <p>In above example, runOnFunction for DominatorTree is called by pass manager
1182 before returning a reference to the desired pass.</p>
1184 <p>
1185 If your pass is capable of updating analyses if they exist (e.g.,
1186 <tt>BreakCriticalEdges</tt>, as described above), you can use the
1187 <tt>getAnalysisIfAvailable</tt> method, which returns a pointer to the analysis
1188 if it is active. For example:</p>
1191 ...
1192 if (DominatorSet *DS = getAnalysisIfAvailable<DominatorSet>()) {
1193 <i>// A DominatorSet is active. This code will update it.</i>
1194 }
1195 ...
1196 </pre></div>
1198 </div>
1200 <!-- *********************************************************************** -->
1203 </div>
1204 <!-- *********************************************************************** -->
1208 <p>Now that we understand the basics of how passes are defined, how they are
1209 used, and how they are required from other passes, it's time to get a little bit
1210 fancier. All of the pass relationships that we have seen so far are very
1211 simple: one pass depends on one other specific pass to be run before it can run.
1212 For many applications, this is great, for others, more flexibility is
1213 required.</p>
1215 <p>In particular, some analyses are defined such that there is a single simple
1216 interface to the analysis results, but multiple ways of calculating them.
1217 Consider alias analysis for example. The most trivial alias analysis returns
1219 flow-sensitive, context-sensitive interprocedural analysis that can take a
1220 significant amount of time to execute (and obviously, there is a lot of room
1221 between these two extremes for other implementations). To cleanly support
1222 situations like this, the LLVM Pass Infrastructure supports the notion of
1223 Analysis Groups.</p>
1225 </div>
1227 <!-- _______________________________________________________________________ -->
1230 </div>
1234 <p>An Analysis Group is a single simple interface that may be implemented by
1235 multiple different passes. Analysis Groups can be given human readable names
1236 just like passes, but unlike passes, they need not derive from the <tt>Pass</tt>
1237 class. An analysis group may have one or more implementations, one of which is
1240 <p>Analysis groups are used by client passes just like other passes are: the
1241 <tt>AnalysisUsage::addRequired()</tt> and <tt>Pass::getAnalysis()</tt> methods.
1243 scans the available passes to see if any implementations of the analysis group
1244 are available. If none is available, the default implementation is created for
1246 between passes</a> still apply.</p>
1249 passes, all analysis group implementations must be registered, and must use the
1251 implementation pool. Also, a default implementation of the interface
1252 <b>must</b> be registered with <A
1255 <p>As a concrete example of an Analysis Group in action, consider the <a
1257 analysis group. The default implementation of the alias analysis interface (the
1258 <tt><a
1260 pass) just does a few simple checks that don't require significant analysis to
1261 compute (such as: two different globals can never alias each other, etc).
1262 Passes that use the <tt><a
1264 interface (for example the <tt><a
1266 not care which implementation of alias analysis is actually provided, they just
1267 use the designated interface.</p>
1269 <p>From the user's perspective, commands work just like normal. Issuing the
1270 command '<tt>opt -gcse ...</tt>' will cause the <tt>basicaa</tt> class to be
1271 instantiated and added to the pass sequence. Issuing the command '<tt>opt
1272 -somefancyaa -gcse ...</tt>' will cause the <tt>gcse</tt> pass to use the
1273 <tt>somefancyaa</tt> alias analysis (which doesn't actually exist, it's just a
1274 hypothetical example) instead.</p>
1276 </div>
1278 <!-- _______________________________________________________________________ -->
1281 </div>
1285 <p>The <tt>RegisterAnalysisGroup</tt> template is used to register the analysis
1286 group itself, while the <tt>INITIALIZE_AG_PASS</tt> is used to add pass
1287 implementations to the analysis group. First,
1288 an analysis group should be registered, with a human readable name
1289 provided for it.
1290 Unlike registration of passes, there is no command line argument to be specified
1294 <b>static</b> RegisterAnalysisGroup<<a href="http://llvm.org/doxygen/classllvm_1_1AliasAnalysis.html">AliasAnalysis</a>> A("<i>Alias Analysis</i>");
1295 </pre></div>
1297 <p>Once the analysis is registered, passes can declare that they are valid
1298 implementations of the interface by using the following code:</p>
1301 <b>namespace</b> {
1302 //<i> Declare that we implement the AliasAnalysis interface</i>
1303 INITIALIZE_AG_PASS(FancyAA, <a href="http://llvm.org/doxygen/classllvm_1_1AliasAnalysis.html">AliasAnalysis</a>, "<i>somefancyaa</i>",
1305 false, // <i>Is CFG Only?</i>
1306 true, // <i>Is Analysis?</i>
1307 false, // <i>Is default Analysis Group implementation?</i>
1308 );
1309 }
1310 </pre></div>
1312 <p>This just shows a class <tt>FancyAA</tt> that
1313 uses the <tt>INITIALIZE_AG_PASS</tt> macro both to register and
1314 to "join" the <tt><a href="http://llvm.org/doxygen/classllvm_1_1AliasAnalysis.html">AliasAnalysis</a></tt>
1315 analysis group. Every implementation of an analysis group should join using
1316 this macro.</p>
1319 <b>namespace</b> {
1320 //<i> Declare that we implement the AliasAnalysis interface</i>
1321 INITIALIZE_AG_PASS(BasicAA, <a href="http://llvm.org/doxygen/classllvm_1_1AliasAnalysis.html">AliasAnalysis</a>, "<i>basicaa</i>",
1323 false, // <i>Is CFG Only?</i>
1324 true, // <i>Is Analysis?</i>
1325 true, // <i>Is default Analysis Group implementation?</i>
1326 );
1327 }
1328 </pre></div>
1330 <p>Here we show how the default implementation is specified (using the final
1331 argument to the <tt>INITIALIZE_AG_PASS</tt> template). There must be exactly
1332 one default implementation available at all times for an Analysis Group to be
1333 used. Only default implementation can derive from <tt>ImmutablePass</tt>.
1334 Here we declare that the
1335 <tt><a href="http://llvm.org/doxygen/structBasicAliasAnalysis.html">BasicAliasAnalysis</a></tt>
1336 pass is the default implementation for the interface.</p>
1338 </div>
1340 <!-- *********************************************************************** -->
1343 </div>
1344 <!-- *********************************************************************** -->
1347 <p>The <a
1349 class is designed to be an easy way to expose various success
1350 metrics from passes. These statistics are printed at the end of a
1351 run, when the -stats command line option is enabled on the command
1352 line. See the <a href="http://llvm.org/docs/ProgrammersManual.html#Statistic">Statistics section</a> in the Programmer's Manual for details.
1354 </div>
1357 <!-- *********************************************************************** -->
1360 </div>
1361 <!-- *********************************************************************** -->
1365 <p>The <a
1367 <a
1370 are set up correctly, and then schedules passes to run efficiently. All of the
1371 LLVM tools that run passes use the <tt>PassManager</tt> for execution of these
1372 passes.</p>
1374 <p>The <tt>PassManager</tt> does two main things to try to reduce the execution
1375 time of a series of passes:</p>
1377 <ol>
1378 <li><b>Share analysis results</b> - The PassManager attempts to avoid
1379 recomputing analysis results as much as possible. This means keeping track of
1380 which analyses are available already, which analyses get invalidated, and which
1381 analyses are needed to be run for a pass. An important part of work is that the
1382 <tt>PassManager</tt> tracks the exact lifetime of all analysis results, allowing
1384 results as soon as they are no longer needed.</li>
1386 <li><b>Pipeline the execution of passes on the program</b> - The
1387 <tt>PassManager</tt> attempts to get better cache and memory usage behavior out
1388 of a series of passes by pipelining the passes together. This means that, given
1391 the first function, then all of the <a
1393 etc... until the entire program has been run through the passes.
1395 <p>This improves the cache behavior of the compiler, because it is only touching
1396 the LLVM program representation for a single function at a time, instead of
1397 traversing the entire program. It reduces the memory consumption of compiler,
1398 because, for example, only one <a
1400 needs to be calculated at a time. This also makes it possible to implement
1401 some <a
1404 </ol>
1406 <p>The effectiveness of the <tt>PassManager</tt> is influenced directly by how
1407 much information it has about the behaviors of the passes it is scheduling. For
1410 Not implementing when it should be implemented will have the effect of not
1411 allowing any analysis results to live across the execution of your pass.</p>
1413 <p>The <tt>PassManager</tt> class exposes a <tt>--debug-pass</tt> command line
1414 options that is useful for debugging pass execution, seeing how things work, and
1415 diagnosing when you should be preserving more analyses than you currently are
1416 (To get information about all of the variants of the <tt>--debug-pass</tt>
1417 option, just type '<tt>opt -help-hidden</tt>').</p>
1419 <p>By using the <tt>--debug-pass=Structure</tt> option, for example, we can see
1421 Lets try it out with the <tt>gcse</tt> and <tt>licm</tt> passes:</p>
1424 $ opt -load ../../../Debug+Asserts/lib/Hello.so -gcse -licm --debug-pass=Structure < hello.bc > /dev/null
1425 Module Pass Manager
1426 Function Pass Manager
1427 Dominator Set Construction
1428 Immediate Dominators Construction
1429 Global Common Subexpression Elimination
1430 -- Immediate Dominators Construction
1431 -- Global Common Subexpression Elimination
1432 Natural Loop Construction
1433 Loop Invariant Code Motion
1434 -- Natural Loop Construction
1435 -- Loop Invariant Code Motion
1436 Module Verifier
1437 -- Dominator Set Construction
1438 -- Module Verifier
1439 Bitcode Writer
1440 --Bitcode Writer
1441 </pre></div>
1443 <p>This output shows us when passes are constructed and when the analysis
1444 results are known to be dead (prefixed with '<tt>--</tt>'). Here we see that
1445 GCSE uses dominator and immediate dominator information to do its job. The LICM
1446 pass uses natural loop information, which uses dominator sets, but not immediate
1447 dominators. Because immediate dominators are no longer useful after the GCSE
1448 pass, it is immediately destroyed. The dominator sets are then reused to
1449 compute natural loop information, which is then used by the LICM pass.</p>
1451 <p>After the LICM pass, the module verifier runs (which is automatically added
1452 by the '<tt>opt</tt>' tool), which uses the dominator set to check that the
1453 resultant LLVM code is well formed. After it finishes, the dominator set
1454 information is destroyed, after being computed once, and shared by three
1455 passes.</p>
1458 World</a> pass in between the two passes:</p>
1461 $ opt -load ../../../Debug+Asserts/lib/Hello.so -gcse -hello -licm --debug-pass=Structure < hello.bc > /dev/null
1462 Module Pass Manager
1463 Function Pass Manager
1464 Dominator Set Construction
1465 Immediate Dominators Construction
1466 Global Common Subexpression Elimination
1467 <b>-- Dominator Set Construction</b>
1468 -- Immediate Dominators Construction
1469 -- Global Common Subexpression Elimination
1470 <b> Hello World Pass
1471 -- Hello World Pass
1472 Dominator Set Construction</b>
1473 Natural Loop Construction
1474 Loop Invariant Code Motion
1475 -- Natural Loop Construction
1476 -- Loop Invariant Code Motion
1477 Module Verifier
1478 -- Dominator Set Construction
1479 -- Module Verifier
1480 Bitcode Writer
1481 --Bitcode Writer
1482 Hello: __main
1483 Hello: puts
1484 Hello: main
1485 </pre></div>
1488 Dominator Set pass, even though it doesn't modify the code at all! To fix this,
1489 we need to add the following <a
1493 <i>// We don't modify the program, so we preserve all analyses</i>
1494 <b>virtual void</b> getAnalysisUsage(AnalysisUsage &AU) <b>const</b> {
1495 AU.setPreservesAll();
1496 }
1497 </pre></div>
1499 <p>Now when we run our pass, we get this output:</p>
1502 $ opt -load ../../../Debug+Asserts/lib/Hello.so -gcse -hello -licm --debug-pass=Structure < hello.bc > /dev/null
1503 Pass Arguments: -gcse -hello -licm
1504 Module Pass Manager
1505 Function Pass Manager
1506 Dominator Set Construction
1507 Immediate Dominators Construction
1508 Global Common Subexpression Elimination
1509 -- Immediate Dominators Construction
1510 -- Global Common Subexpression Elimination
1511 Hello World Pass
1512 -- Hello World Pass
1513 Natural Loop Construction
1514 Loop Invariant Code Motion
1515 -- Loop Invariant Code Motion
1516 -- Natural Loop Construction
1517 Module Verifier
1518 -- Dominator Set Construction
1519 -- Module Verifier
1520 Bitcode Writer
1521 --Bitcode Writer
1522 Hello: __main
1523 Hello: puts
1524 Hello: main
1525 </pre></div>
1527 <p>Which shows that we don't accidentally invalidate dominator information
1528 anymore, and therefore do not have to compute it twice.</p>
1530 </div>
1532 <!-- _______________________________________________________________________ -->
1535 </div>
1540 <b>virtual void</b> releaseMemory();
1541 </pre></div>
1543 <p>The <tt>PassManager</tt> automatically determines when to compute analysis
1544 results, and how long to keep them around for. Because the lifetime of the pass
1545 object itself is effectively the entire duration of the compilation process, we
1546 need some way to free analysis results when they are no longer useful. The
1547 <tt>releaseMemory</tt> virtual method is the way to do this.</p>
1549 <p>If you are writing an analysis or any other pass that retains a significant
1552 <tt>releaseMemory</tt> to, well, release the memory allocated to maintain this
1553 internal state. This method is called after the <tt>run*</tt> method for the
1554 class, before the next call of <tt>run*</tt> in your pass.</p>
1556 </div>
1558 <!-- *********************************************************************** -->
1561 </div>
1562 <!-- *********************************************************************** -->
1566 <p><i>Size matters</i> when constructing production quality tools using llvm,
1567 both for the purposes of distribution, and for regulating the resident code size
1568 when running on the target system. Therefore, it becomes desirable to
1569 selectively use some passes, while omitting others and maintain the flexibility
1570 to change configurations later on. You want to be able to do all this, and,
1571 provide feedback to the user. This is where pass registration comes into
1572 play.</p>
1574 <p>The fundamental mechanisms for pass registration are the
1575 <tt>MachinePassRegistry</tt> class and subclasses of
1576 <tt>MachinePassRegistryNode</tt>.</p>
1578 <p>An instance of <tt>MachinePassRegistry</tt> is used to maintain a list of
1579 <tt>MachinePassRegistryNode</tt> objects. This instance maintains the list and
1580 communicates additions and deletions to the command line interface.</p>
1582 <p>An instance of <tt>MachinePassRegistryNode</tt> subclass is used to maintain
1583 information provided about a particular pass. This information includes the
1584 command line name, the command help string and the address of the function used
1585 to create an instance of the pass. A global static constructor of one of these
1586 instances <i>registers</i> with a corresponding <tt>MachinePassRegistry</tt>,
1587 the static destructor <i>unregisters</i>. Thus a pass that is statically linked
1588 in the tool will be registered at start up. A dynamically loaded pass will
1589 register on load and unregister at unload.</p>
1591 </div>
1593 <!-- _______________________________________________________________________ -->
1596 </div>
1600 <p>There are predefined registries to track instruction scheduling
1601 (<tt>RegisterScheduler</tt>) and register allocation (<tt>RegisterRegAlloc</tt>)
1602 machine passes. Here we will describe how to <i>register</i> a register
1603 allocator machine pass.</p>
1605 <p>Implement your register allocator machine pass. In your register allocator
1606 .cpp file add the following include;</p>
1610 </pre></div>
1612 <p>Also in your register allocator .cpp file, define a creator function in the
1613 form; </p>
1616 FunctionPass *createMyRegisterAllocator() {
1617 return new MyRegisterAllocator();
1618 }
1619 </pre></div>
1621 <p>Note that the signature of this function should match the type of
1622 <tt>RegisterRegAlloc::FunctionPassCtor</tt>. In the same file add the
1628 createMyRegisterAllocator);
1629 </pre></div>
1631 <p>Note the two spaces prior to the help string produces a tidy result on the
1632 -help query.</p>
1635 $ llc -help
1636 ...
1637 -regalloc - Register allocator to use (default=linearscan)
1638 =linearscan - linear scan register allocator
1639 =local - local register allocator
1640 =simple - simple register allocator
1641 =myregalloc - my register allocator help string
1642 ...
1643 </pre></div>
1645 <p>And that's it. The user is now free to use <tt>-regalloc=myregalloc</tt> as
1646 an option. Registering instruction schedulers is similar except use the
1647 <tt>RegisterScheduler</tt> class. Note that the
1648 <tt>RegisterScheduler::FunctionPassCtor</tt> is significantly different from
1649 <tt>RegisterRegAlloc::FunctionPassCtor</tt>.</p>
1651 <p>To force the load/linking of your register allocator into the llc/lli tools,
1653 call line to <tt>llvm/Codegen/LinkAllCodegenComponents.h</tt>.</p>
1655 </div>
1658 <!-- _______________________________________________________________________ -->
1661 </div>
1665 <p>The easiest way to get started is to clone one of the existing registries; we
1666 recommend <tt>llvm/CodeGen/RegAllocRegistry.h</tt>. The key things to modify
1667 are the class name and the <tt>FunctionPassCtor</tt> type.</p>
1669 <p>Then you need to declare the registry. Example: if your pass registry is
1670 <tt>RegisterMyPasses</tt> then define;</p>
1673 MachinePassRegistry RegisterMyPasses::Registry;
1674 </pre></div>
1676 <p>And finally, declare the command line option for your passes. Example:</p>
1679 cl::opt<RegisterMyPasses::FunctionPassCtor, false,
1680 RegisterPassParser<RegisterMyPasses> >
1682 cl::init(&createDefaultMyPass),
1684 </pre></div>
1687 creator.</p>
1689 </div>
1691 <!-- *********************************************************************** -->
1694 </div>
1695 <!-- *********************************************************************** -->
1699 <p>Unfortunately, using GDB with dynamically loaded passes is not as easy as it
1700 should be. First of all, you can't set a breakpoint in a shared object that has
1701 not been loaded yet, and second of all there are problems with inlined functions
1702 in shared objects. Here are some suggestions to debugging your pass with
1703 GDB.</p>
1705 <p>For sake of discussion, I'm going to assume that you are debugging a
1706 transformation invoked by <tt>opt</tt>, although nothing described here depends
1707 on that.</p>
1709 </div>
1711 <!-- _______________________________________________________________________ -->
1714 </div>
1718 <p>First thing you do is start <tt>gdb</tt> on the <tt>opt</tt> process:</p>
1721 $ <b>gdb opt</b>
1722 GNU gdb 5.0
1723 Copyright 2000 Free Software Foundation, Inc.
1724 GDB is free software, covered by the GNU General Public License, and you are
1725 welcome to change it and/or distribute copies of it under certain conditions.
1729 (gdb)
1730 </pre></div>
1732 <p>Note that <tt>opt</tt> has a lot of debugging information in it, so it takes
1733 time to load. Be patient. Since we cannot set a breakpoint in our pass yet
1734 (the shared object isn't loaded until runtime), we must execute the process, and
1735 have it stop before it invokes our pass, but after it has loaded the shared
1736 object. The most foolproof way of doing this is to set a breakpoint in
1737 <tt>PassManager::run</tt> and then run the process with the arguments you
1738 want:</p>
1741 (gdb) <b>break llvm::PassManager::run</b>
1742 Breakpoint 1 at 0x2413bc: file Pass.cpp, line 70.
1743 (gdb) <b>run test.bc -load $(LLVMTOP)/llvm/Debug+Asserts/lib/[libname].so -[passoption]</b>
1744 Starting program: opt test.bc -load $(LLVMTOP)/llvm/Debug+Asserts/lib/[libname].so -[passoption]
1745 Breakpoint 1, PassManager::run (this=0xffbef174, M=@0x70b298) at Pass.cpp:70
1746 70 bool PassManager::run(Module &M) { return PM->run(M); }
1747 (gdb)
1748 </pre></div>
1750 <p>Once the <tt>opt</tt> stops in the <tt>PassManager::run</tt> method you are
1751 now free to set breakpoints in your pass so that you can trace through execution
1752 or do other standard debugging stuff.</p>
1754 </div>
1756 <!-- _______________________________________________________________________ -->
1759 </div>
1763 <p>Once you have the basics down, there are a couple of problems that GDB has,
1764 some with solutions, some without.</p>
1766 <ul>
1767 <li>Inline functions have bogus stack information. In general, GDB does a
1768 pretty good job getting stack traces and stepping through inline functions.
1769 When a pass is dynamically loaded however, it somehow completely loses this
1770 capability. The only solution I know of is to de-inline a function (move it
1771 from the body of a class to a .cpp file).</li>
1773 <li>Restarting the program breaks breakpoints. After following the information
1774 above, you have succeeded in getting some breakpoints planted in your pass. Nex
1775 thing you know, you restart the program (i.e., you type '<tt>run</tt>' again),
1776 and you start getting errors about breakpoints being unsettable. The only way I
1778 already set in your pass, run the program, and re-set the breakpoints once
1779 execution stops in <tt>PassManager::run</tt>.</li>
1781 </ul>
1783 <p>Hopefully these tips will help with common case debugging situations. If
1784 you'd like to contribute some tips of your own, just contact <a
1787 </div>
1789 <!-- *********************************************************************** -->
1792 </div>
1793 <!-- *********************************************************************** -->
1797 <p>Although the LLVM Pass Infrastructure is very capable as it stands, and does
1798 some nifty stuff, there are things we'd like to add in the future. Here is
1799 where we are going:</p>
1801 </div>
1803 <!-- _______________________________________________________________________ -->
1806 </div>
1810 <p>Multiple CPU machines are becoming more common and compilation can never be
1811 fast enough: obviously we should allow for a multithreaded compiler. Because of
1812 the semantics defined for passes above (specifically they cannot maintain state
1813 across invocations of their <tt>run*</tt> methods), a nice clean way to
1814 implement a multithreaded compiler would be for the <tt>PassManager</tt> class
1815 to create multiple instances of each pass object, and allow the separate
1816 instances to be hacking on different parts of the program at the same time.</p>
1818 <p>This implementation would prevent each of the passes from having to implement
1819 multithreaded constructs, requiring only the LLVM core to have locking in a few
1820 places (for global resources). Although this is a simple extension, we simply
1821 haven't had time (or multiprocessor machines, thus a reason) to implement this.
1822 Despite that, we have kept the LLVM passes SMP ready, and you should too.</p>
1824 </div>
1826 <!-- *********************************************************************** -->
1827 <hr>
1828 <address>
1836 Last modified: $Date$
1837 </address>
1839 </body>
1840 </html>