docs/Bugpoint.html

   1 <!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01//EN"
   2                       "http://www.w3.org/TR/html4/strict.dtd">
   3 <html>
   4 <head>
   5   <title>LLVM bugpoint tool: design and usage</title>
   6   <link rel="stylesheet" href="llvm.css" type="text/css">
   7 </head>
   8
   9 <div class="doc_title">
  10   LLVM bugpoint tool: design and usage
  11 </div>
  12
  13 <ul>
  14   <li><a href="#desc">Description</a></li>
  15   <li><a href="#design">Design Philosophy</a>
  16   <ul>
  17     <li><a href="#autoselect">Automatic Debugger Selection</a></li>
  18     <li><a href="#crashdebug">Crash debugger</a></li>
  19     <li><a href="#codegendebug">Code generator debugger</a></li>
  20     <li><a href="#miscompilationdebug">Miscompilation debugger</a></li>
  21   </ul></li>
  22   <li><a href="#advice">Advice for using <tt>bugpoint</tt></a></li>
  23 </ul>
  24
  25 <div class="doc_author">
  26 <p>Written by <a href="mailto:sabre@nondot.org">Chris Lattner</a></p>
  27 </div>
  28
  29 <!-- *********************************************************************** -->
  30 <div class="doc_section">
  31 <a name="desc">Description</a>
  32 </div>
  33 <!-- *********************************************************************** -->
  34
  35 <div class="doc_text">
  36
  37 <p><tt>bugpoint</tt> narrows down the source of problems in LLVM tools and
  38 passes.  It can be used to debug three types of failures: optimizer crashes,
  39 miscompilations by optimizers, or bad native code generation (including problems
  40 in the static and JIT compilers).  It aims to reduce large test cases to small,
  41 useful ones.  For example, if <tt>opt</tt> crashes while optimizing a
  42 file, it will identify the optimization (or combination of optimizations) that
  43 causes the crash, and reduce the file down to a small example which triggers the
  44 crash.</p>
  45
  46 <p>For detailed case scenarios, such as debugging <tt>opt</tt>,
  47 <tt>llvm-ld</tt>, or one of the LLVM code generators, see <a
  48 href="HowToSubmitABug.html">How To Submit a Bug Report document</a>.</p>
  49
  50 </div>
  51
  52 <!-- *********************************************************************** -->
  53 <div class="doc_section">
  54 <a name="design">Design Philosophy</a>
  55 </div>
  56 <!-- *********************************************************************** -->
  57
  58 <div class="doc_text">
  59
  60 <p><tt>bugpoint</tt> is designed to be a useful tool without requiring any
  61 hooks into the LLVM infrastructure at all.  It works with any and all LLVM
  62 passes and code generators, and does not need to "know" how they work.  Because
  63 of this, it may appear to do stupid things or miss obvious
  64 simplifications.  <tt>bugpoint</tt> is also designed to trade off programmer
  65 time for computer time in the compiler-debugging process; consequently, it may
  66 take a long period of (unattended) time to reduce a test case, but we feel it
  67 is still worth it. Note that <tt>bugpoint</tt> is generally very quick unless
  68 debugging a miscompilation where each test of the program (which requires
  69 executing it) takes a long time.</p>
  70
  71 </div>
  72
  73 <!-- ======================================================================= -->
  74 <div class="doc_subsection">
  75   <a name="autoselect">Automatic Debugger Selection</a>
  76 </div>
  77
  78 <div class="doc_text">
  79
  80 <p><tt>bugpoint</tt> reads each <tt>.bc</tt> or <tt>.ll</tt> file specified on
  81 the command line and links them together into a single module, called the test
  82 program.  If any LLVM passes are specified on the command line, it runs these
  83 passes on the test program.  If any of the passes crash, or if they produce
  84 malformed output (which causes the verifier to abort), <tt>bugpoint</tt> starts
  85 the <a href="#crashdebug">crash debugger</a>.</p>
  86
  87 <p>Otherwise, if the <tt>-output</tt> option was not specified,
  88 <tt>bugpoint</tt> runs the test program with the C backend (which is assumed to
  89 generate good code) to generate a reference output.  Once <tt>bugpoint</tt> has
  90 a reference output for the test program, it tries executing it with the
  91 selected code generator.  If the selected code generator crashes,
  92 <tt>bugpoint</tt> starts the <a href="#crashdebug">crash debugger</a> on the
  93 code generator.  Otherwise, if the resulting output differs from the reference
  94 output, it assumes the difference resulted from a code generator failure, and
  95 starts the <a href="#codegendebug">code generator debugger</a>.</p>
  96
  97 <p>Finally, if the output of the selected code generator matches the reference
  98 output, <tt>bugpoint</tt> runs the test program after all of the LLVM passes
  99 have been applied to it.  If its output differs from the reference output, it
 100 assumes the difference resulted from a failure in one of the LLVM passes, and
 101 enters the <a href="#miscompilationdebug">miscompilation debugger</a>.
 102 Otherwise, there is no problem <tt>bugpoint</tt> can debug.</p>
 103
 104 </div>
 105
 106 <!-- ======================================================================= -->
 107 <div class="doc_subsection">
 108   <a name="crashdebug">Crash debugger</a>
 109 </div>
 110
 111 <div class="doc_text">
 112
 113 <p>If an optimizer or code generator crashes, <tt>bugpoint</tt> will try as hard
 114 as it can to reduce the list of passes (for optimizer crashes) and the size of
 115 the test program.  First, <tt>bugpoint</tt> figures out which combination of
 116 optimizer passes triggers the bug. This is useful when debugging a problem
 117 exposed by <tt>opt</tt>, for example, because it runs over 38 passes.</p>
 118
 119 <p>Next, <tt>bugpoint</tt> tries removing functions from the test program, to
 120 reduce its size.  Usually it is able to reduce a test program to a single
 121 function, when debugging intraprocedural optimizations.  Once the number of
 122 functions has been reduced, it attempts to delete various edges in the control
 123 flow graph, to reduce the size of the function as much as possible.  Finally,
 124 <tt>bugpoint</tt> deletes any individual LLVM instructions whose absence does
 125 not eliminate the failure.  At the end, <tt>bugpoint</tt> should tell you what
 126 passes crash, give you a bitcode file, and give you instructions on how to
 127 reproduce the failure with <tt>opt</tt> or <tt>llc</tt>.</p>
 128
 129 </div>
 130
 131 <!-- ======================================================================= -->
 132 <div class="doc_subsection">
 133   <a name="codegendebug">Code generator debugger</a>
 134 </div>
 135
 136 <div class="doc_text">
 137
 138 <p>The code generator debugger attempts to narrow down the amount of code that
 139 is being miscompiled by the selected code generator.  To do this, it takes the
 140 test program and partitions it into two pieces: one piece which it compiles
 141 with the C backend (into a shared object), and one piece which it runs with
 142 either the JIT or the static LLC compiler.  It uses several techniques to
 143 reduce the amount of code pushed through the LLVM code generator, to reduce the
 144 potential scope of the problem.  After it is finished, it emits two bitcode
 145 files (called "test" [to be compiled with the code generator] and "safe" [to be
 146 compiled with the C backend], respectively), and instructions for reproducing
 147 the problem.  The code generator debugger assumes that the C backend produces
 148 good code.</p>
 149
 150 </div>
 151
 152 <!-- ======================================================================= -->
 153 <div class="doc_subsection">
 154   <a name="miscompilationdebug">Miscompilation debugger</a>
 155 </div>
 156
 157 <div class="doc_text">
 158
 159 <p>The miscompilation debugger works similarly to the code generator debugger.
 160 It works by splitting the test program into two pieces, running the
 161 optimizations specified on one piece, linking the two pieces back together, and
 162 then executing the result.  It attempts to narrow down the list of passes to
 163 the one (or few) which are causing the miscompilation, then reduce the portion
 164 of the test program which is being miscompiled.  The miscompilation debugger
 165 assumes that the selected code generator is working properly.</p>
 166
 167 </div>
 168
 169 <!-- *********************************************************************** -->
 170 <div class="doc_section">
 171   <a name="advice">Advice for using bugpoint</a>
 172 </div>
 173 <!-- *********************************************************************** -->
 174
 175 <div class="doc_text">
 176
 177 <tt>bugpoint</tt> can be a remarkably useful tool, but it sometimes works in
 178 non-obvious ways.  Here are some hints and tips:<p>
 179
 180 <ol>
 181 <li>In the code generator and miscompilation debuggers, <tt>bugpoint</tt> only
 182     works with programs that have deterministic output.  Thus, if the program
 183     outputs <tt>argv[0]</tt>, the date, time, or any other "random" data,
 184     <tt>bugpoint</tt> may misinterpret differences in these data, when output,
 185     as the result of a miscompilation.  Programs should be temporarily modified
 186     to disable outputs that are likely to vary from run to run.
 187
 188 <li>In the code generator and miscompilation debuggers, debugging will go
 189     faster if you manually modify the program or its inputs to reduce the
 190     runtime, but still exhibit the problem.
 191
 192 <li><tt>bugpoint</tt> is extremely useful when working on a new optimization:
 193     it helps track down regressions quickly.  To avoid having to relink
 194     <tt>bugpoint</tt> every time you change your optimization however, have
 195     <tt>bugpoint</tt> dynamically load your optimization with the
 196     <tt>-load</tt> option.
 197
 198 <li><p><tt>bugpoint</tt> can generate a lot of output and run for a long period
 199     of time.  It is often useful to capture the output of the program to file.
 200     For example, in the C shell, you can run:</p>
 201
 202 <div class="doc_code">
 203 <p><tt>bugpoint  ... |&amp; tee bugpoint.log</tt></p>
 204 </div>
 205
 206     <p>to get a copy of <tt>bugpoint</tt>'s output in the file
 207     <tt>bugpoint.log</tt>, as well as on your terminal.</p>
 208
 209 <li><tt>bugpoint</tt> cannot debug problems with the LLVM linker. If
 210     <tt>bugpoint</tt> crashes before you see its "All input ok" message,
 211     you might try <tt>llvm-link -v</tt> on the same set of input files. If
 212     that also crashes, you may be experiencing a linker bug.
 213
 214 <li><tt>bugpoint</tt> is useful for proactively finding bugs in LLVM.
 215     Invoking <tt>bugpoint</tt> with the <tt>-find-bugs</tt> option will cause
 216     the list of specified optimizations to be randomized and applied to the
 217     program. This process will repeat until a bug is found or the user
 218     kills <tt>bugpoint</tt>.
 219
 220 <li><p><tt>bugpoint</tt> does not understand the <tt>-O</tt> option
 221     that is used to specify optimization level to <tt>opt</tt>. You
 222     can use e.g.</p>
 223
 224 <div class="doc_code">
 225 <p><tt>opt -O2 -debug-pass=Arguments foo.bc -disable-output</tt></p>
 226 </div>
 227
 228     <p>to get a list of passes that are used with <tt>-O2</tt> and
 229     then pass this list to <tt>bugpoint</tt>.</p>
 230
 231 </ol>
 232
 233 </div>
 234
 235 <!-- *********************************************************************** -->
 236
 237 <hr>
 238 <address>
 239   <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
 240   src="http://jigsaw.w3.org/css-validator/images/vcss-blue" alt="Valid CSS"></a>
 241   <a href="http://validator.w3.org/check/referer"><img
 242   src="http://www.w3.org/Icons/valid-html401-blue" alt="Valid HTML 4.01"></a>
 243
 244   <a href="mailto:sabre@nondot.org">Chris Lattner</a><br>
 245   <a href="http://llvm.org">LLVM Compiler Infrastructure</a><br>
 246   Last modified: $Date$
 247 </address>
 248
 249 </body>
 250 </html>