1 <?xml version="1.0" encoding="iso-8859-1"?>
2 <sect1 id="options-debugging">
3 <title>Debugging the compiler</title>
5 <indexterm><primary>debugging options (for GHC)</primary></indexterm>
7 <para>HACKER TERRITORY. HACKER TERRITORY. (You were warned.)</para>
9 <sect2 id="dumping-output">
10 <title>Dumping out compiler intermediate structures</title>
12 <indexterm><primary>dumping GHC intermediates</primary></indexterm>
13 <indexterm><primary>intermediate passes, output</primary></indexterm>
18 <option>-ddump-</option><replaceable>pass</replaceable>
19 <indexterm><primary><option>-ddump</option> options</primary></indexterm>
22 <para>Make a debugging dump after pass
23 <literal><pass></literal> (may be common enough to need
24 a short form…). You can get all of these at once
25 (<emphasis>lots</emphasis> of output) by using
26 <option>-v5</option>, or most of them with
27 <option>-v4</option>. Some of the most useful ones
33 <option>-ddump-parsed</option>:
34 <indexterm><primary><option>-ddump-parsed</option></primary></indexterm>
37 <para>parser output</para>
43 <option>-ddump-rn</option>:
44 <indexterm><primary><option>-ddump-rn</option></primary></indexterm>
47 <para>renamer output</para>
53 <option>-ddump-tc</option>:
54 <indexterm><primary><option>-ddump-tc</option></primary></indexterm>
57 <para>typechecker output</para>
63 <option>-ddump-splices</option>:
64 <indexterm><primary><option>-ddump-splices</option></primary></indexterm>
67 <para>Dump Template Haskell expressions that we splice in,
68 and what Haskell code the expression evaluates to.</para>
74 <option>-ddump-types</option>:
75 <indexterm><primary><option>-ddump-types</option></primary></indexterm>
78 <para>Dump a type signature for each value defined at
79 the top level of the module. The list is sorted
80 alphabetically. Using <option>-dppr-debug</option>
81 dumps a type signature for all the imported and
82 system-defined things as well; useful for debugging the
89 <option>-ddump-deriv</option>:
90 <indexterm><primary><option>-ddump-deriv</option></primary></indexterm>
93 <para>derived instances</para>
99 <option>-ddump-ds</option>:
100 <indexterm><primary><option>-ddump-ds</option></primary></indexterm>
103 <para>desugarer output</para>
109 <option>-ddump-spec</option>:
110 <indexterm><primary><option>-ddump-spec</option></primary></indexterm>
113 <para>output of specialisation pass</para>
119 <option>-ddump-rules</option>:
120 <indexterm><primary><option>-ddump-rules</option></primary></indexterm>
123 <para>dumps all rewrite rules (including those generated
124 by the specialisation pass)</para>
130 <option>-ddump-simpl</option>:
131 <indexterm><primary><option>-ddump-simpl</option></primary></indexterm>
134 <para>simplifier output (Core-to-Core passes)</para>
140 <option>-ddump-inlinings</option>:
141 <indexterm><primary><option>-ddump-inlinings</option></primary></indexterm>
144 <para>inlining info from the simplifier</para>
150 <option>-ddump-cpranal</option>:
151 <indexterm><primary><option>-ddump-cpranal</option></primary></indexterm>
154 <para>CPR analyser output</para>
160 <option>-ddump-stranal</option>:
161 <indexterm><primary><option>-ddump-stranal</option></primary></indexterm>
164 <para>strictness analyser output</para>
170 <option>-ddump-cse</option>:
171 <indexterm><primary><option>-ddump-cse</option></primary></indexterm>
174 <para>CSE pass output</para>
180 <option>-ddump-workwrap</option>:
181 <indexterm><primary><option>-ddump-workwrap</option></primary></indexterm>
184 <para>worker/wrapper split output</para>
190 <option>-ddump-occur-anal</option>:
191 <indexterm><primary><option>-ddump-occur-anal</option></primary></indexterm>
194 <para>`occurrence analysis' output</para>
200 <option>-ddump-prep</option>:
201 <indexterm><primary><option>-ddump-prep</option></primary></indexterm>
204 <para>output of core preparation pass</para>
210 <option>-ddump-stg</option>:
211 <indexterm><primary><option>-ddump-stg</option></primary></indexterm>
214 <para>output of STG-to-STG passes</para>
220 <option>-ddump-flatC</option>:
221 <indexterm><primary><option>-ddump-flatC</option></primary></indexterm>
224 <para><emphasis>flattened</emphasis> Abstract C</para>
230 <option>-ddump-cmm</option>:
231 <indexterm><primary><option>-ddump-cmm</option></primary></indexterm>
234 <para>Print the C-- code out.</para>
240 <option>-ddump-opt-cmm</option>:
241 <indexterm><primary><option>-ddump-opt-cmm</option></primary></indexterm>
244 <para>Dump the results of C-- to C-- optimising passes.</para>
250 <option>-ddump-asm</option>:
251 <indexterm><primary><option>-ddump-asm</option></primary></indexterm>
254 <para>assembly language from the native-code generator</para>
260 <option>-ddump-bcos</option>:
261 <indexterm><primary><option>-ddump-bcos</option></primary></indexterm>
264 <para>byte code compiler output</para>
270 <option>-ddump-foreign</option>:
271 <indexterm><primary><option>-ddump-foreign</option></primary></indexterm>
274 <para>dump foreign export stubs</para>
283 <option>-ddump-simpl-iterations</option>:
284 <indexterm><primary><option>-ddump-simpl-iterations</option></primary></indexterm>
287 <para>Show the output of each <emphasis>iteration</emphasis>
288 of the simplifier (each run of the simplifier has a maximum
289 number of iterations, normally 4). Used when even
290 <option>-dverbose-simpl</option> doesn't cut it.</para>
296 <option>-ddump-simpl-stats</option>
297 <indexterm><primary><option>-ddump-simpl-stats option</option></primary></indexterm>
300 <para>Dump statistics about how many of each kind of
301 transformation too place. If you add
302 <option>-dppr-debug</option> you get more detailed
309 <option>-ddump-if-trace</option>
310 <indexterm><primary><option>-ddump-if-trace</option></primary></indexterm>
313 <para>Make the interface loader be *real* chatty about what it is
320 <option>-ddump-tc-trace</option>
321 <indexterm><primary><option>-ddump-tc-trace</option></primary></indexterm>
324 <para>Make the type checker be *real* chatty about what it is
331 <option>-ddump-rn-trace</option>
332 <indexterm><primary><option>-ddump-rn-trace</option></primary></indexterm>
335 <para>Make the renamer be *real* chatty about what it is
342 <option>-ddump-rn-stats</option>
343 <indexterm><primary><option>-dshow-rn-stats</option></primary></indexterm>
346 <para>Print out summary of what kind of information the renamer
347 had to bring in.</para>
353 <option>-dverbose-core2core</option>
354 <indexterm><primary><option>-dverbose-core2core</option></primary></indexterm>
357 <option>-dverbose-stg2stg</option>
358 <indexterm><primary><option>-dverbose-stg2stg</option></primary></indexterm>
361 <para>Show the output of the intermediate Core-to-Core and
362 STG-to-STG passes, respectively. (<emphasis>Lots</emphasis>
363 of output!) So: when we're really desperate:</para>
366 % ghc -noC -O -ddump-simpl -dverbose-simpl -dcore-lint Foo.hs
374 <option>-dshow-passes</option>
375 <indexterm><primary><option>-dshow-passes</option></primary></indexterm>
378 <para>Print out each pass name as it happens.</para>
384 <option>-dfaststring-stats</option>
385 <indexterm><primary><option>-dfaststring-stats</option></primary></indexterm>
388 <para>Show statistics for the usage of fast strings by the
395 <option>-dppr-debug</option>
396 <indexterm><primary><option>-dppr-debug</option></primary></indexterm>
399 <para>Debugging output is in one of several
400 “styles.” Take the printing of types, for
401 example. In the “user” style (the default), the
402 compiler's internal ideas about types are presented in
403 Haskell source-level syntax, insofar as possible. In the
404 “debug” style (which is the default for
405 debugging output), the types are printed in with explicit
406 foralls, and variables have their unique-id attached (so you
407 can check for things that look the same but aren't). This
408 flag makes debugging output appear in the more verbose debug
415 <option>-dppr-user-length</option>
416 <indexterm><primary><option>-dppr-user-length</option></primary></indexterm>
419 <para>In error messages, expressions are printed to a
420 certain “depth”, with subexpressions beyond the
421 depth replaced by ellipses. This flag sets the
422 depth. Its default value is 5.</para>
428 <option>-dshow-unused-imports</option>
429 <indexterm><primary><option>-dshow-unused-imports</option></primary></indexterm>
432 <para>Have the renamer report what imports does not
439 <sect2 id="checking-consistency">
440 <title>Checking for consistency</title>
442 <indexterm><primary>consistency checks</primary></indexterm>
443 <indexterm><primary>lint</primary></indexterm>
449 <option>-dcore-lint</option>
450 <indexterm><primary><option>-dcore-lint</option></primary></indexterm>
453 <para>Turn on heavyweight intra-pass sanity-checking within
454 GHC, at Core level. (It checks GHC's sanity, not yours.)</para>
460 <option>-dstg-lint</option>:
461 <indexterm><primary><option>-dstg-lint</option></primary></indexterm>
464 <para>Ditto for STG level. (NOTE: currently doesn't work).</para>
470 <option>-dcmm-lint</option>:
471 <indexterm><primary><option>-dcmm-lint</option></primary></indexterm>
474 <para>Ditto for C-- level.</para>
482 <title>How to read Core syntax (from some <option>-ddump</option>
485 <indexterm><primary>reading Core syntax</primary></indexterm>
486 <indexterm><primary>Core syntax, how to read</primary></indexterm>
488 <para>Let's do this by commenting an example. It's from doing
489 <option>-ddump-ds</option> on this code:
492 skip2 m = m : skip2 (m+2)
495 Before we jump in, a word about names of things. Within GHC,
496 variables, type constructors, etc., are identified by their
497 “Uniques.” These are of the form `letter' plus
498 `number' (both loosely interpreted). The `letter' gives some idea
499 of where the Unique came from; e.g., <literal>_</literal>
500 means “built-in type variable”; <literal>t</literal>
501 means “from the typechecker”; <literal>s</literal>
502 means “from the simplifier”; and so on. The `number'
503 is printed fairly compactly in a `base-62' format, which everyone
504 hates except me (WDP).</para>
506 <para>Remember, everything has a “Unique” and it is
507 usually printed out when debugging, in some form or another. So
508 here we go…</para>
512 Main.skip2{-r1L6-} :: _forall_ a$_4 =>{{Num a$_4}} -> a$_4 -> [a$_4]
514 --# `r1L6' is the Unique for Main.skip2;
515 --# `_4' is the Unique for the type-variable (template) `a'
516 --# `{{Num a$_4}}' is a dictionary argument
520 --# `_NI_' means "no (pragmatic) information" yet; it will later
521 --# evolve into the GHC_PRAGMA info that goes into interface files.
524 /\ _4 -> \ d.Num.t4Gt ->
527 +.t4Hg :: _4 -> _4 -> _4
529 +.t4Hg = (+{-r3JH-} _4) d.Num.t4Gt
531 fromInt.t4GS :: Int{-2i-} -> _4
533 fromInt.t4GS = (fromInt{-r3JX-} _4) d.Num.t4Gt
535 --# The `+' class method (Unique: r3JH) selects the addition code
536 --# from a `Num' dictionary (now an explicit lambda'd argument).
537 --# Because Core is 2nd-order lambda-calculus, type applications
538 --# and lambdas (/\) are explicit. So `+' is first applied to a
539 --# type (`_4'), then to a dictionary, yielding the actual addition
540 --# function that we will use subsequently...
542 --# We play the exact same game with the (non-standard) class method
543 --# `fromInt'. Unsurprisingly, the type `Int' is wired into the
553 } in fromInt.t4GS ds.d4Qz
555 --# `I# 2#' is just the literal Int `2'; it reflects the fact that
556 --# GHC defines `data Int = I# Int#', where Int# is the primitive
557 --# unboxed type. (see relevant info about unboxed types elsewhere...)
559 --# The `!' after `I#' indicates that this is a *saturated*
560 --# application of the `I#' data constructor (i.e., not partially
563 skip2.t3Ja :: _4 -> [_4]
567 let { ds.d4QQ :: [_4]
573 ds.d4QY = +.t4Hg m.r1H4 lit.t4Hb
574 } in skip2.t3Ja ds.d4QY
582 <para>(“It's just a simple functional language” is an
583 unregisterised trademark of Peyton Jones Enterprises, plc.)</para>
588 <title>Unregisterised compilation</title>
589 <indexterm><primary>unregisterised compilation</primary></indexterm>
591 <para>The term "unregisterised" really means "compile via vanilla
592 C", disabling some of the platform-specific tricks that GHC
593 normally uses to make programs go faster. When compiling
594 unregisterised, GHC simply generates a C file which is compiled
597 <para>Unregisterised compilation can be useful when porting GHC to
598 a new machine, since it reduces the prerequisite tools to
599 <command>gcc</command>, <command>as</command>, and
600 <command>ld</command> and nothing more, and furthermore the amount
601 of platform-specific code that needs to be written in order to get
602 unregisterised compilation going is usually fairly small.</para>
604 <para>Unregisterised compilation cannot be selected at
605 compile-time; you have to build GHC with the appropriate options
606 set. Consult the GHC Building Guide for details.</para>
612 ;;; Local Variables: ***
614 ;;; sgml-parent-document: ("users_guide.xml" "book" "chapter" "sect1") ***