Home | History | Annotate | Download | only in docs
      1 ====================================
      2 LLVM bugpoint tool: design and usage
      3 ====================================
      4 
      5 .. contents::
      6    :local:
      7 
      8 Description
      9 ===========
     10 
     11 ``bugpoint`` narrows down the source of problems in LLVM tools and passes.  It
     12 can be used to debug three types of failures: optimizer crashes, miscompilations
     13 by optimizers, or bad native code generation (including problems in the static
     14 and JIT compilers).  It aims to reduce large test cases to small, useful ones.
     15 For example, if ``opt`` crashes while optimizing a file, it will identify the
     16 optimization (or combination of optimizations) that causes the crash, and reduce
     17 the file down to a small example which triggers the crash.
     18 
     19 For detailed case scenarios, such as debugging ``opt``, or one of the LLVM code
     20 generators, see `How To Submit a Bug Report document <HowToSubmitABug.html>`_.
     21 
     22 Design Philosophy
     23 =================
     24 
     25 ``bugpoint`` is designed to be a useful tool without requiring any hooks into
     26 the LLVM infrastructure at all.  It works with any and all LLVM passes and code
     27 generators, and does not need to "know" how they work.  Because of this, it may
     28 appear to do stupid things or miss obvious simplifications.  ``bugpoint`` is
     29 also designed to trade off programmer time for computer time in the
     30 compiler-debugging process; consequently, it may take a long period of
     31 (unattended) time to reduce a test case, but we feel it is still worth it. Note
     32 that ``bugpoint`` is generally very quick unless debugging a miscompilation
     33 where each test of the program (which requires executing it) takes a long time.
     34 
     35 Automatic Debugger Selection
     36 ----------------------------
     37 
     38 ``bugpoint`` reads each ``.bc`` or ``.ll`` file specified on the command line
     39 and links them together into a single module, called the test program.  If any
     40 LLVM passes are specified on the command line, it runs these passes on the test
     41 program.  If any of the passes crash, or if they produce malformed output (which
     42 causes the verifier to abort), ``bugpoint`` starts the `crash debugger`_.
     43 
     44 Otherwise, if the ``-output`` option was not specified, ``bugpoint`` runs the
     45 test program with the "safe" backend (which is assumed to generate good code) to
     46 generate a reference output.  Once ``bugpoint`` has a reference output for the
     47 test program, it tries executing it with the selected code generator.  If the
     48 selected code generator crashes, ``bugpoint`` starts the `crash debugger`_ on
     49 the code generator.  Otherwise, if the resulting output differs from the
     50 reference output, it assumes the difference resulted from a code generator
     51 failure, and starts the `code generator debugger`_.
     52 
     53 Finally, if the output of the selected code generator matches the reference
     54 output, ``bugpoint`` runs the test program after all of the LLVM passes have
     55 been applied to it.  If its output differs from the reference output, it assumes
     56 the difference resulted from a failure in one of the LLVM passes, and enters the
     57 `miscompilation debugger`_.  Otherwise, there is no problem ``bugpoint`` can
     58 debug.
     59 
     60 .. _crash debugger:
     61 
     62 Crash debugger
     63 --------------
     64 
     65 If an optimizer or code generator crashes, ``bugpoint`` will try as hard as it
     66 can to reduce the list of passes (for optimizer crashes) and the size of the
     67 test program.  First, ``bugpoint`` figures out which combination of optimizer
     68 passes triggers the bug. This is useful when debugging a problem exposed by
     69 ``opt``, for example, because it runs over 38 passes.
     70 
     71 Next, ``bugpoint`` tries removing functions from the test program, to reduce its
     72 size.  Usually it is able to reduce a test program to a single function, when
     73 debugging intraprocedural optimizations.  Once the number of functions has been
     74 reduced, it attempts to delete various edges in the control flow graph, to
     75 reduce the size of the function as much as possible.  Finally, ``bugpoint``
     76 deletes any individual LLVM instructions whose absence does not eliminate the
     77 failure.  At the end, ``bugpoint`` should tell you what passes crash, give you a
     78 bitcode file, and give you instructions on how to reproduce the failure with
     79 ``opt`` or ``llc``.
     80 
     81 .. _code generator debugger:
     82 
     83 Code generator debugger
     84 -----------------------
     85 
     86 The code generator debugger attempts to narrow down the amount of code that is
     87 being miscompiled by the selected code generator.  To do this, it takes the test
     88 program and partitions it into two pieces: one piece which it compiles with the
     89 "safe" backend (into a shared object), and one piece which it runs with either
     90 the JIT or the static LLC compiler.  It uses several techniques to reduce the
     91 amount of code pushed through the LLVM code generator, to reduce the potential
     92 scope of the problem.  After it is finished, it emits two bitcode files (called
     93 "test" [to be compiled with the code generator] and "safe" [to be compiled with
     94 the "safe" backend], respectively), and instructions for reproducing the
     95 problem.  The code generator debugger assumes that the "safe" backend produces
     96 good code.
     97 
     98 .. _miscompilation debugger:
     99 
    100 Miscompilation debugger
    101 -----------------------
    102 
    103 The miscompilation debugger works similarly to the code generator debugger.  It
    104 works by splitting the test program into two pieces, running the optimizations
    105 specified on one piece, linking the two pieces back together, and then executing
    106 the result.  It attempts to narrow down the list of passes to the one (or few)
    107 which are causing the miscompilation, then reduce the portion of the test
    108 program which is being miscompiled.  The miscompilation debugger assumes that
    109 the selected code generator is working properly.
    110 
    111 Advice for using bugpoint
    112 =========================
    113 
    114 ``bugpoint`` can be a remarkably useful tool, but it sometimes works in
    115 non-obvious ways.  Here are some hints and tips:
    116 
    117 * In the code generator and miscompilation debuggers, ``bugpoint`` only works
    118   with programs that have deterministic output.  Thus, if the program outputs
    119   ``argv[0]``, the date, time, or any other "random" data, ``bugpoint`` may
    120   misinterpret differences in these data, when output, as the result of a
    121   miscompilation.  Programs should be temporarily modified to disable outputs
    122   that are likely to vary from run to run.
    123 
    124 * In the code generator and miscompilation debuggers, debugging will go faster
    125   if you manually modify the program or its inputs to reduce the runtime, but
    126   still exhibit the problem.
    127 
    128 * ``bugpoint`` is extremely useful when working on a new optimization: it helps
    129   track down regressions quickly.  To avoid having to relink ``bugpoint`` every
    130   time you change your optimization however, have ``bugpoint`` dynamically load
    131   your optimization with the ``-load`` option.
    132 
    133 * ``bugpoint`` can generate a lot of output and run for a long period of time.
    134   It is often useful to capture the output of the program to file.  For example,
    135   in the C shell, you can run:
    136 
    137   .. code-block:: console
    138 
    139     $ bugpoint  ... |& tee bugpoint.log
    140 
    141   to get a copy of ``bugpoint``'s output in the file ``bugpoint.log``, as well
    142   as on your terminal.
    143 
    144 * ``bugpoint`` cannot debug problems with the LLVM linker. If ``bugpoint``
    145   crashes before you see its "All input ok" message, you might try ``llvm-link
    146   -v`` on the same set of input files. If that also crashes, you may be
    147   experiencing a linker bug.
    148 
    149 * ``bugpoint`` is useful for proactively finding bugs in LLVM.  Invoking
    150   ``bugpoint`` with the ``-find-bugs`` option will cause the list of specified
    151   optimizations to be randomized and applied to the program. This process will
    152   repeat until a bug is found or the user kills ``bugpoint``.
    153 
    154 What to do when bugpoint isn't enough
    155 =====================================
    156 	
    157 Sometimes, ``bugpoint`` is not enough. In particular, InstCombine and
    158 TargetLowering both have visitor structured code with lots of potential
    159 transformations.  If the process of using bugpoint has left you with still too
    160 much code to figure out and the problem seems to be in instcombine, the
    161 following steps may help.  These same techniques are useful with TargetLowering
    162 as well.
    163 
    164 Turn on ``-debug-only=instcombine`` and see which transformations within
    165 instcombine are firing by selecting out lines with "``IC``" in them.
    166 
    167 At this point, you have a decision to make.  Is the number of transformations
    168 small enough to step through them using a debugger?  If so, then try that.
    169 
    170 If there are too many transformations, then a source modification approach may
    171 be helpful.  In this approach, you can modify the source code of instcombine to
    172 disable just those transformations that are being performed on your test input
    173 and perform a binary search over the set of transformations.  One set of places
    174 to modify are the "``visit*``" methods of ``InstCombiner`` (*e.g.*
    175 ``visitICmpInst``) by adding a "``return false``" as the first line of the
    176 method.
    177 
    178 If that still doesn't remove enough, then change the caller of
    179 ``InstCombiner::DoOneIteration``, ``InstCombiner::runOnFunction`` to limit the
    180 number of iterations.
    181 
    182 You may also find it useful to use "``-stats``" now to see what parts of
    183 instcombine are firing.  This can guide where to put additional reporting code.
    184 
    185 At this point, if the amount of transformations is still too large, then
    186 inserting code to limit whether or not to execute the body of the code in the
    187 visit function can be helpful.  Add a static counter which is incremented on
    188 every invocation of the function.  Then add code which simply returns false on
    189 desired ranges.  For example:
    190 
    191 .. code-block:: c++
    192 
    193 
    194   static int calledCount = 0;
    195   calledCount++;
    196   DEBUG(if (calledCount < 212) return false);
    197   DEBUG(if (calledCount > 217) return false);
    198   DEBUG(if (calledCount == 213) return false);
    199   DEBUG(if (calledCount == 214) return false);
    200   DEBUG(if (calledCount == 215) return false);
    201   DEBUG(if (calledCount == 216) return false);
    202   DEBUG(dbgs() << "visitXOR calledCount: " << calledCount << "\n");
    203   DEBUG(dbgs() << "I: "; I->dump());
    204 
    205 could be added to ``visitXOR`` to limit ``visitXor`` to being applied only to
    206 calls 212 and 217. This is from an actual test case and raises an important
    207 point---a simple binary search may not be sufficient, as transformations that
    208 interact may require isolating more than one call.  In TargetLowering, use
    209 ``return SDNode();`` instead of ``return false;``.
    210 
    211 Now that that the number of transformations is down to a manageable number, try
    212 examining the output to see if you can figure out which transformations are
    213 being done.  If that can be figured out, then do the usual debugging.  If which
    214 code corresponds to the transformation being performed isn't obvious, set a
    215 breakpoint after the call count based disabling and step through the code.
    216 Alternatively, you can use "``printf``" style debugging to report waypoints.
    217