Merge branch 'master' of darcs.haskell.org:/home/darcs/ghc
[ghc.git] / docs / users_guide / using.xml
1 <?xml version="1.0" encoding="iso-8859-1"?>
2 <chapter id="using-ghc">
3 <title>Using GHC</title>
4
5 <indexterm><primary>GHC, using</primary></indexterm>
6 <indexterm><primary>using GHC</primary></indexterm>
7
8 <sect1>
9 <title>Getting started: compiling programs</title>
10
11 <para>
12 In this chapter you'll find a complete reference to the GHC
13 command-line syntax, including all 400+ flags. It's a large and
14 complex system, and there are lots of details, so it can be
15 quite hard to figure out how to get started. With that in mind,
16 this introductory section provides a quick introduction to the
17 basic usage of GHC for compiling a Haskell program, before the
18 following sections dive into the full syntax.
19 </para>
20
21 <para>
22 Let's create a Hello World program, and compile and run it.
23 First, create a file <filename>hello.hs</filename> containing
24 the Haskell code:
25 </para>
26
27 <programlisting>
28 main = putStrLn "Hello, World!"
29 </programlisting>
30
31 <para>To compile the program, use GHC like this:</para>
32
33 <screen>
34 $ ghc hello.hs
35 </screen>
36
37 <para>(where <literal>$</literal> represents the prompt: don't
38 type it). GHC will compile the source
39 file <filename>hello.hs</filename>, producing
40 an <firstterm>object
41 file</firstterm> <filename>hello.o</filename> and
42 an <firstterm>interface
43 file</firstterm> <filename>hello.hi</filename>, and then it
44 will link the object file to the libraries that come with GHC
45 to produce an executable called <filename>hello</filename> on
46 Unix/Linux/Mac, or <filename>hello.exe</filename> on
47 Windows.</para>
48
49 <para>
50 By default GHC will be very quiet about what it is doing, only
51 printing error messages. If you want to see in more detail
52 what's going on behind the scenes, add <option>-v</option> to
53 the command line.
54 </para>
55
56 <para>
57 Then we can run the program like this:
58 </para>
59
60 <screen>
61 $ ./hello
62 Hello World!
63 </screen>
64
65 <para>
66 If your program contains multiple modules, then you only need to
67 tell GHC the name of the source file containing
68 the <filename>Main</filename> module, and GHC will examine
69 the <literal>import</literal> declarations to find the other
70 modules that make up the program and find their source files.
71 This means that, with the exception of
72 the <literal>Main</literal> module, every source file should be
73 named after the module name that it contains (with dots replaced
74 by directory separators). For example, the
75 module <literal>Data.Person</literal> would be in the
76 file <filename>Data/Person.hs</filename> on Unix/Linux/Mac,
77 or <filename>Data\Person.hs</filename> on Windows.
78 </para>
79 </sect1>
80
81 <sect1>
82 <title>Options overview</title>
83
84 <para>GHC's behaviour is controlled by
85 <firstterm>options</firstterm>, which for historical reasons are
86 also sometimes referred to as command-line flags or arguments.
87 Options can be specified in three ways:</para>
88
89 <sect2>
90 <title>Command-line arguments</title>
91
92 <indexterm><primary>structure, command-line</primary></indexterm>
93 <indexterm><primary>command-line</primary><secondary>arguments</secondary></indexterm>
94 <indexterm><primary>arguments</primary><secondary>command-line</secondary></indexterm>
95
96 <para>An invocation of GHC takes the following form:</para>
97
98 <screen>
99 ghc [argument...]
100 </screen>
101
102 <para>Command-line arguments are either options or file names.</para>
103
104 <para>Command-line options begin with <literal>-</literal>.
105 They may <emphasis>not</emphasis> be grouped:
106 <option>-vO</option> is different from <option>-v -O</option>.
107 Options need not precede filenames: e.g., <literal>ghc *.o -o
108 foo</literal>. All options are processed and then applied to
109 all files; you cannot, for example, invoke <literal>ghc -c -O1
110 Foo.hs -O2 Bar.hs</literal> to apply different optimisation
111 levels to the files <filename>Foo.hs</filename> and
112 <filename>Bar.hs</filename>.</para>
113 </sect2>
114
115 <sect2 id="source-file-options">
116 <title>Command line options in source files</title>
117
118 <indexterm><primary>source-file options</primary></indexterm>
119
120 <para>Sometimes it is useful to make the connection between a
121 source file and the command-line options it requires quite
122 tight. For instance, if a Haskell source file deliberately
123 uses name shadowing, it should be compiled with the
124 <option>-fno-warn-name-shadowing</option> option. Rather than maintaining
125 the list of per-file options in a <filename>Makefile</filename>,
126 it is possible to do this directly in the source file using the
127 <literal>OPTIONS_GHC</literal> pragma <indexterm><primary>OPTIONS_GHC
128 pragma</primary></indexterm>:</para>
129
130 <programlisting>
131 {-# OPTIONS_GHC -fno-warn-name-shadowing #-}
132 module X where
133 ...
134 </programlisting>
135
136 <para><literal>OPTIONS_GHC</literal> is a <emphasis>file-header pragma</emphasis>
137 (see <xref linkend="pragmas"/>).</para>
138
139 <para>Only <emphasis>dynamic</emphasis> flags can be used in an <literal>OPTIONS_GHC</literal> pragma
140 (see <xref linkend="static-dynamic-flags"/>).</para>
141
142 <para>Note that your command shell does not
143 get to the source file options, they are just included literally
144 in the array of command-line arguments the compiler
145 maintains internally, so you'll be desperately disappointed if
146 you try to glob etc. inside <literal>OPTIONS_GHC</literal>.</para>
147
148 <para>NOTE: the contents of OPTIONS_GHC are appended to the
149 command-line options, so options given in the source file
150 override those given on the command-line.</para>
151
152 <para>It is not recommended to move all the contents of your
153 Makefiles into your source files, but in some circumstances, the
154 <literal>OPTIONS_GHC</literal> pragma is the Right Thing. (If you
155 use <option>-keep-hc-file</option> and have OPTION flags in
156 your module, the OPTIONS_GHC will get put into the generated .hc
157 file).</para>
158 </sect2>
159
160 <sect2>
161 <title>Setting options in GHCi</title>
162
163 <para>Options may also be modified from within GHCi, using the
164 <literal>:set</literal> command. See <xref linkend="ghci-set"/>
165 for more details.</para>
166 </sect2>
167 </sect1>
168
169 <sect1 id="static-dynamic-flags">
170 <title>Static, Dynamic, and Mode options</title>
171 <indexterm><primary>static</primary><secondary>options</secondary>
172 </indexterm>
173 <indexterm><primary>dynamic</primary><secondary>options</secondary>
174 </indexterm>
175 <indexterm><primary>mode</primary><secondary>options</secondary>
176 </indexterm>
177
178 <para>Each of GHC's command line options is classified as
179 <firstterm>static</firstterm>, <firstterm>dynamic</firstterm> or
180 <firstterm>mode</firstterm>:</para>
181
182 <variablelist>
183 <varlistentry>
184 <term>Mode flags</term>
185 <listitem>
186 <para>For example, <option>&ndash;&ndash;make</option> or <option>-E</option>.
187 There may only be a single mode flag on the command line. The
188 available modes are listed in <xref linkend="modes"/>.</para>
189 </listitem>
190 </varlistentry>
191 <varlistentry>
192 <term>Dynamic Flags</term>
193 <listitem>
194 <para>Most non-mode flags fall into this category. A dynamic flag
195 may be used on the command line, in a
196 <literal>OPTIONS_GHC</literal> pragma in a source file, or set
197 using <literal>:set</literal> in GHCi.</para>
198 </listitem>
199 </varlistentry>
200 <varlistentry>
201 <term>Static Flags</term>
202 <listitem>
203 <para>A few flags are "static", which means they can only be used on
204 the command-line, and remain in force over the entire GHC/GHCi
205 run.</para>
206 </listitem>
207 </varlistentry>
208 </variablelist>
209
210 <para>The flag reference tables (<xref
211 linkend="flag-reference"/>) lists the status of each flag.</para>
212
213 <para>There are a few flags that are static except that they can
214 also be used with GHCi's <literal>:set</literal> command; these
215 are listed as &ldquo;static/<literal>:set</literal>&rdquo; in the
216 table.</para>
217 </sect1>
218
219 <sect1 id="file-suffixes">
220 <title>Meaningful file suffixes</title>
221
222 <indexterm><primary>suffixes, file</primary></indexterm>
223 <indexterm><primary>file suffixes for GHC</primary></indexterm>
224
225 <para>File names with &ldquo;meaningful&rdquo; suffixes (e.g.,
226 <filename>.lhs</filename> or <filename>.o</filename>) cause the
227 &ldquo;right thing&rdquo; to happen to those files.</para>
228
229 <variablelist>
230
231 <varlistentry>
232 <term><filename>.hs</filename></term>
233 <listitem>
234 <para>A Haskell module.</para>
235 </listitem>
236 </varlistentry>
237
238 <varlistentry>
239 <term>
240 <filename>.lhs</filename>
241 <indexterm><primary><literal>lhs</literal> suffix</primary></indexterm>
242 </term>
243 <listitem>
244 <para>A &ldquo;literate Haskell&rdquo; module.</para>
245 </listitem>
246 </varlistentry>
247
248 <varlistentry>
249 <term><filename>.hi</filename></term>
250 <listitem>
251 <para>A Haskell interface file, probably
252 compiler-generated.</para>
253 </listitem>
254 </varlistentry>
255
256 <varlistentry>
257 <term><filename>.hc</filename></term>
258 <listitem>
259 <para>Intermediate C file produced by the Haskell
260 compiler.</para>
261 </listitem>
262 </varlistentry>
263
264 <varlistentry>
265 <term><filename>.c</filename></term>
266 <listitem>
267 <para>A C&nbsp;file not produced by the Haskell
268 compiler.</para>
269 </listitem>
270 </varlistentry>
271
272 <varlistentry>
273 <term><filename>.ll</filename></term>
274 <listitem>
275 <para>An llvm-intermediate-language source file, usually
276 produced by the compiler.</para>
277 </listitem>
278 </varlistentry>
279
280 <varlistentry>
281 <term><filename>.bc</filename></term>
282 <listitem>
283 <para>An llvm-intermediate-language bitcode file, usually
284 produced by the compiler.</para>
285 </listitem>
286 </varlistentry>
287
288 <varlistentry>
289 <term><filename>.s</filename></term>
290 <listitem>
291 <para>An assembly-language source file, usually produced by
292 the compiler.</para>
293 </listitem>
294 </varlistentry>
295
296 <varlistentry>
297 <term><filename>.o</filename></term>
298 <listitem>
299 <para>An object file, produced by an assembler.</para>
300 </listitem>
301 </varlistentry>
302 </variablelist>
303
304 <para>Files with other suffixes (or without suffixes) are passed
305 straight to the linker.</para>
306
307 </sect1>
308
309 <sect1 id="modes">
310 <title>Modes of operation</title>
311
312 <para>
313 GHC's behaviour is firstly controlled by a mode flag. Only one
314 of these flags may be given, but it does not necessarily need to
315 be the first option on the command-line.
316 </para>
317
318 <para>
319 If no mode flag is present, then GHC will enter make mode
320 (<xref linkend="make-mode" />) if there are any Haskell source
321 files given on the command line, or else it will link the
322 objects named on the command line to produce an executable.
323 </para>
324
325 <para>The available mode flags are:</para>
326
327 <variablelist>
328 <varlistentry>
329 <term>
330 <cmdsynopsis><command>ghc --interactive</command>
331 </cmdsynopsis>
332 <indexterm><primary>interactive mode</primary></indexterm>
333 <indexterm><primary>ghci</primary></indexterm>
334 </term>
335 <listitem>
336 <para>Interactive mode, which is also available as
337 <command>ghci</command>. Interactive mode is described in
338 more detail in <xref linkend="ghci"/>.</para>
339 </listitem>
340 </varlistentry>
341
342 <varlistentry>
343 <term>
344 <cmdsynopsis><command>ghc &ndash;&ndash;make</command>
345 </cmdsynopsis>
346 <indexterm><primary>make mode</primary></indexterm>
347 <indexterm><primary><option>&ndash;&ndash;make</option></primary></indexterm>
348 </term>
349 <listitem>
350 <para>In this mode, GHC will build a multi-module Haskell
351 program automatically, figuring out dependencies for itself.
352 If you have a straightforward Haskell program, this is
353 likely to be much easier, and faster, than using
354 <command>make</command>. Make mode is described in <xref
355 linkend="make-mode"/>.</para>
356
357 <para>
358 This mode is the default if there are any Haskell
359 source files mentioned on the command line, and in this case
360 the <option>&ndash;&ndash;make</option> option can be omitted.
361 </para>
362 </listitem>
363 </varlistentry>
364
365 <varlistentry>
366 <term>
367 <cmdsynopsis><command>ghc -e</command>
368 <arg choice='plain'><replaceable>expr</replaceable></arg>
369 </cmdsynopsis>
370 <indexterm><primary>eval mode</primary></indexterm>
371 </term>
372 <listitem>
373 <para>Expression-evaluation mode. This is very similar to
374 interactive mode, except that there is a single expression
375 to evaluate (<replaceable>expr</replaceable>) which is given
376 on the command line. See <xref linkend="eval-mode"/> for
377 more details.</para>
378 </listitem>
379 </varlistentry>
380
381 <varlistentry>
382 <term>
383 <cmdsynopsis>
384 <command>ghc -E</command>
385 <command>ghc -c</command>
386 <command>ghc -S</command>
387 <command>ghc -c</command>
388 </cmdsynopsis>
389 <indexterm><primary><option>-E</option></primary></indexterm>
390 <indexterm><primary><option>-C</option></primary></indexterm>
391 <indexterm><primary><option>-S</option></primary></indexterm>
392 <indexterm><primary><option>-c</option></primary></indexterm>
393 </term>
394 <listitem>
395 <para>This is the traditional batch-compiler mode, in which
396 GHC can compile source files one at a time, or link objects
397 together into an executable. This mode also applies if
398 there is no other mode flag specified on the command line,
399 in which case it means that the specified files should be
400 compiled and then linked to form a program. See <xref
401 linkend="options-order"/>.</para>
402 </listitem>
403 </varlistentry>
404
405 <varlistentry>
406 <term>
407 <cmdsynopsis>
408 <command>ghc -M</command>
409 </cmdsynopsis>
410 <indexterm><primary>dependency-generation mode</primary></indexterm>
411 </term>
412 <listitem>
413 <para>Dependency-generation mode. In this mode, GHC can be
414 used to generate dependency information suitable for use in
415 a <literal>Makefile</literal>. See <xref
416 linkend="makefile-dependencies"/>.</para>
417 </listitem>
418 </varlistentry>
419
420 <varlistentry>
421 <term>
422 <cmdsynopsis>
423 <command>ghc --mk-dll</command>
424 </cmdsynopsis>
425 <indexterm><primary>DLL-creation mode</primary></indexterm>
426 </term>
427 <listitem>
428 <para>DLL-creation mode (Windows only). See <xref
429 linkend="win32-dlls-create"/>.</para>
430 </listitem>
431 </varlistentry>
432
433 <varlistentry>
434 <term>
435 <cmdsynopsis>
436 <command>ghc --help</command> <command>ghc -?</command>
437 </cmdsynopsis>
438 <indexterm><primary><option>&ndash;&ndash;help</option></primary></indexterm>
439 </term>
440 <listitem>
441 <para>Cause GHC to spew a long usage message to standard
442 output and then exit.</para>
443 </listitem>
444 </varlistentry>
445
446 <varlistentry>
447 <term>
448 <cmdsynopsis>
449 <command>ghc --show-iface <replaceable>file</replaceable></command>
450 </cmdsynopsis>
451 <indexterm><primary><option>&ndash;&ndash;--show-iface</option></primary></indexterm>
452 </term>
453 <listitem>
454 <para>Read the interface in
455 <replaceable>file</replaceable> and dump it as text to
456 <literal>stdout</literal>. For example <literal>ghc --show-iface M.hi</literal>.</para>
457 </listitem>
458 </varlistentry>
459
460 <varlistentry>
461 <term>
462 <cmdsynopsis>
463 <command>ghc --supported-extensions</command>
464 <command>ghc --supported-languages</command>
465 </cmdsynopsis>
466 <indexterm><primary><option>&ndash;&ndash;supported-extensions</option></primary><primary><option>&ndash;&ndash;supported-languages</option></primary></indexterm>
467 </term>
468 <listitem>
469 <para>Print the supported language extensions.</para>
470 </listitem>
471 </varlistentry>
472
473 <varlistentry>
474 <term>
475 <cmdsynopsis>
476 <command>ghc --info</command>
477 </cmdsynopsis>
478 <indexterm><primary><option>&ndash;&ndash;info</option></primary></indexterm>
479 </term>
480 <listitem>
481 <para>Print information about the compiler.</para>
482 </listitem>
483 </varlistentry>
484
485 <varlistentry>
486 <term>
487 <cmdsynopsis>
488 <command>ghc --version</command>
489 <command>ghc -V</command>
490 </cmdsynopsis>
491 <indexterm><primary><option>-V</option></primary></indexterm>
492 <indexterm><primary><option>&ndash;&ndash;version</option></primary></indexterm>
493 </term>
494 <listitem>
495 <para>Print a one-line string including GHC's version number.</para>
496 </listitem>
497 </varlistentry>
498
499 <varlistentry>
500 <term>
501 <cmdsynopsis>
502 <command>ghc --numeric-version</command>
503 </cmdsynopsis>
504 <indexterm><primary><option>&ndash;&ndash;numeric-version</option></primary></indexterm>
505 </term>
506 <listitem>
507 <para>Print GHC's numeric version number only.</para>
508 </listitem>
509 </varlistentry>
510
511 <varlistentry>
512 <term>
513 <cmdsynopsis>
514 <command>ghc --print-libdir</command>
515 </cmdsynopsis>
516 <indexterm><primary><option>&ndash;&ndash;print-libdir</option></primary></indexterm>
517 </term>
518 <listitem>
519 <para>Print the path to GHC's library directory. This is
520 the top of the directory tree containing GHC's libraries,
521 interfaces, and include files (usually something like
522 <literal>/usr/local/lib/ghc-5.04</literal> on Unix). This
523 is the value of
524 <literal>$libdir</literal><indexterm><primary><literal>libdir</literal></primary></indexterm>
525 in the package configuration file
526 (see <xref linkend="packages"/>).</para>
527 </listitem>
528 </varlistentry>
529
530 </variablelist>
531
532 <sect2 id="make-mode">
533 <title>Using <command>ghc</command> <option>&ndash;&ndash;make</option></title>
534 <indexterm><primary><option>&ndash;&ndash;make</option></primary></indexterm>
535 <indexterm><primary>separate compilation</primary></indexterm>
536
537 <para>In this mode, GHC will build a multi-module Haskell program by following
538 dependencies from one or more root modules (usually just
539 <literal>Main</literal>). For example, if your
540 <literal>Main</literal> module is in a file called
541 <filename>Main.hs</filename>, you could compile and link the
542 program like this:</para>
543
544 <screen>
545 ghc &ndash;&ndash;make Main.hs
546 </screen>
547
548 <para>
549 In fact, GHC enters make mode automatically if there are any
550 Haskell source files on the command line and no other mode is
551 specified, so in this case we could just type
552 </para>
553
554 <screen>
555 ghc Main.hs
556 </screen>
557
558 <para>Any number of source file names or module names may be
559 specified; GHC will figure out all the modules in the program by
560 following the imports from these initial modules. It will then
561 attempt to compile each module which is out of date, and
562 finally, if there is a <literal>Main</literal> module, the
563 program will also be linked into an executable.</para>
564
565 <para>The main advantages to using <literal>ghc
566 &ndash;&ndash;make</literal> over traditional
567 <literal>Makefile</literal>s are:</para>
568
569 <itemizedlist>
570 <listitem>
571 <para>GHC doesn't have to be restarted for each compilation,
572 which means it can cache information between compilations.
573 Compiling a multi-module program with <literal>ghc
574 &ndash;&ndash;make</literal> can be up to twice as fast as
575 running <literal>ghc</literal> individually on each source
576 file.</para>
577 </listitem>
578 <listitem>
579 <para>You don't have to write a <literal>Makefile</literal>.</para>
580 <indexterm><primary><literal>Makefile</literal>s</primary><secondary>avoiding</secondary></indexterm>
581 </listitem>
582 <listitem>
583 <para>GHC re-calculates the dependencies each time it is
584 invoked, so the dependencies never get out of sync with the
585 source.</para>
586 </listitem>
587 </itemizedlist>
588
589 <para>Any of the command-line options described in the rest of
590 this chapter can be used with
591 <option>&ndash;&ndash;make</option>, but note that any options
592 you give on the command line will apply to all the source files
593 compiled, so if you want any options to apply to a single source
594 file only, you'll need to use an <literal>OPTIONS_GHC</literal>
595 pragma (see <xref linkend="source-file-options"/>).</para>
596
597 <para>If the program needs to be linked with additional objects
598 (say, some auxiliary C code), then the object files can be
599 given on the command line and GHC will include them when linking
600 the executable.</para>
601
602 <para>Note that GHC can only follow dependencies if it has the
603 source file available, so if your program includes a module for
604 which there is no source file, even if you have an object and an
605 interface file for the module, then GHC will complain. The
606 exception to this rule is for package modules, which may or may
607 not have source files.</para>
608
609 <para>The source files for the program don't all need to be in
610 the same directory; the <option>-i</option> option can be used
611 to add directories to the search path (see <xref
612 linkend="search-path"/>).</para>
613 </sect2>
614
615 <sect2 id="eval-mode">
616 <title>Expression evaluation mode</title>
617
618 <para>This mode is very similar to interactive mode, except that
619 there is a single expression to evaluate which is specified on
620 the command line as an argument to the <option>-e</option>
621 option:</para>
622
623 <screen>
624 ghc -e <replaceable>expr</replaceable>
625 </screen>
626
627 <para>Haskell source files may be named on the command line, and
628 they will be loaded exactly as in interactive mode. The
629 expression is evaluated in the context of the loaded
630 modules.</para>
631
632 <para>For example, to load and run a Haskell program containing
633 a module <literal>Main</literal>, we might say</para>
634
635 <screen>
636 ghc -e Main.main Main.hs
637 </screen>
638
639 <para>or we can just use this mode to evaluate expressions in
640 the context of the <literal>Prelude</literal>:</para>
641
642 <screen>
643 $ ghc -e "interact (unlines.map reverse.lines)"
644 hello
645 olleh
646 </screen>
647 </sect2>
648
649 <sect2 id="options-order">
650 <title>Batch compiler mode</title>
651
652 <para>In <emphasis>batch mode</emphasis>, GHC will compile one or more source files
653 given on the command line.</para>
654
655 <para>The first phase to run is determined by each input-file
656 suffix, and the last phase is determined by a flag. If no
657 relevant flag is present, then go all the way through to linking.
658 This table summarises:</para>
659
660 <informaltable>
661 <tgroup cols="4">
662 <colspec align="left"/>
663 <colspec align="left"/>
664 <colspec align="left"/>
665 <colspec align="left"/>
666
667 <thead>
668 <row>
669 <entry>Phase of the compilation system</entry>
670 <entry>Suffix saying &ldquo;start here&rdquo;</entry>
671 <entry>Flag saying &ldquo;stop after&rdquo;</entry>
672 <entry>(suffix of) output file</entry>
673 </row>
674 </thead>
675 <tbody>
676 <row>
677 <entry>literate pre-processor</entry>
678 <entry><literal>.lhs</literal></entry>
679 <entry>-</entry>
680 <entry><literal>.hs</literal></entry>
681 </row>
682
683 <row>
684 <entry>C pre-processor (opt.) </entry>
685 <entry><literal>.hs</literal> (with
686 <option>-cpp</option>)</entry>
687 <entry><option>-E</option></entry>
688 <entry><literal>.hspp</literal></entry>
689 </row>
690
691 <row>
692 <entry>Haskell compiler</entry>
693 <entry><literal>.hs</literal></entry>
694 <entry><option>-C</option>, <option>-S</option></entry>
695 <entry><literal>.hc</literal>, <literal>.s</literal></entry>
696 </row>
697
698 <row>
699 <entry>C compiler (opt.)</entry>
700 <entry><literal>.hc</literal> or <literal>.c</literal></entry>
701 <entry><option>-S</option></entry>
702 <entry><literal>.s</literal></entry>
703 </row>
704
705 <row>
706 <entry>assembler</entry>
707 <entry><literal>.s</literal></entry>
708 <entry><option>-c</option></entry>
709 <entry><literal>.o</literal></entry>
710 </row>
711
712 <row>
713 <entry>linker</entry>
714 <entry><replaceable>other</replaceable></entry>
715 <entry>-</entry>
716 <entry><filename>a.out</filename></entry>
717 </row>
718 </tbody>
719 </tgroup>
720 </informaltable>
721
722 <indexterm><primary><option>-C</option></primary></indexterm>
723 <indexterm><primary><option>-E</option></primary></indexterm>
724 <indexterm><primary><option>-S</option></primary></indexterm>
725 <indexterm><primary><option>-c</option></primary></indexterm>
726
727 <para>Thus, a common invocation would be: </para>
728
729 <screen>
730 ghc -c Foo.hs
731 </screen>
732
733 <para>to compile the Haskell source file
734 <filename>Foo.hs</filename> to an object file
735 <filename>Foo.o</filename>.</para>
736
737 <para>Note: What the Haskell compiler proper produces depends on what
738 backend code generator is used. See <xref linkend="code-generators"/>
739 for more details.</para>
740
741 <para>Note: C pre-processing is optional, the
742 <option>-cpp</option><indexterm><primary><option>-cpp</option></primary></indexterm>
743 flag turns it on. See <xref linkend="c-pre-processor"/> for more
744 details.</para>
745
746 <para>Note: The option <option>-E</option><indexterm><primary>-E
747 option</primary></indexterm> runs just the pre-processing passes
748 of the compiler, dumping the result in a file.</para>
749
750 <sect3 id="overriding-suffixes">
751 <title>Overriding the default behaviour for a file</title>
752
753 <para>As described above, the way in which a file is processed by GHC
754 depends on its suffix. This behaviour can be overridden using the
755 <option>-x</option> option:</para>
756
757 <variablelist>
758 <varlistentry>
759 <term><option>-x</option> <replaceable>suffix</replaceable>
760 <indexterm><primary><option>-x</option></primary>
761 </indexterm></term>
762 <listitem>
763 <para>Causes all files following this option on the command
764 line to be processed as if they had the suffix
765 <replaceable>suffix</replaceable>. For example, to compile a
766 Haskell module in the file <literal>M.my-hs</literal>,
767 use <literal>ghc -c -x hs M.my-hs</literal>.</para>
768 </listitem>
769 </varlistentry>
770 </variablelist>
771 </sect3>
772
773 </sect2>
774 </sect1>
775
776 <sect1 id="options-help">
777 <title>Help and verbosity options</title>
778
779 <indexterm><primary>help options</primary></indexterm>
780 <indexterm><primary>verbosity options</primary></indexterm>
781
782 <para>See also the <option>--help</option>, <option>--version</option>, <option>--numeric-version</option>,
783 and <option>--print-libdir</option> modes in <xref linkend="modes"/>.</para>
784 <variablelist>
785 <varlistentry>
786 <term>
787 <option>-v</option>
788 <indexterm><primary><option>-v</option></primary></indexterm>
789 </term>
790 <listitem>
791 <para>The <option>-v</option> option makes GHC
792 <emphasis>verbose</emphasis>: it reports its version number
793 and shows (on stderr) exactly how it invokes each phase of
794 the compilation system. Moreover, it passes the
795 <option>-v</option> flag to most phases; each reports its
796 version number (and possibly some other information).</para>
797
798 <para>Please, oh please, use the <option>-v</option> option
799 when reporting bugs! Knowing that you ran the right bits in
800 the right order is always the first thing we want to
801 verify.</para>
802 </listitem>
803 </varlistentry>
804
805 <varlistentry>
806 <term>
807 <option>-v</option><replaceable>n</replaceable>
808 <indexterm><primary><option>-v</option></primary></indexterm>
809 </term>
810 <listitem>
811 <para>To provide more control over the compiler's verbosity,
812 the <option>-v</option> flag takes an optional numeric
813 argument. Specifying <option>-v</option> on its own is
814 equivalent to <option>-v3</option>, and the other levels
815 have the following meanings:</para>
816
817 <variablelist>
818 <varlistentry>
819 <term><option>-v0</option></term>
820 <listitem>
821 <para>Disable all non-essential messages (this is the
822 default).</para>
823 </listitem>
824 </varlistentry>
825
826 <varlistentry>
827 <term><option>-v1</option></term>
828 <listitem>
829 <para>Minimal verbosity: print one line per
830 compilation (this is the default when
831 <option>&ndash;&ndash;make</option> or
832 <option>&ndash;&ndash;interactive</option> is on).</para>
833 </listitem>
834 </varlistentry>
835
836 <varlistentry>
837 <term><option>-v2</option></term>
838 <listitem>
839 <para>Print the name of each compilation phase as it
840 is executed. (equivalent to
841 <option>-dshow-passes</option>).</para>
842 </listitem>
843 </varlistentry>
844
845 <varlistentry>
846 <term><option>-v3</option></term>
847 <listitem>
848 <para>The same as <option>-v2</option>, except that in
849 addition the full command line (if appropriate) for
850 each compilation phase is also printed.</para>
851 </listitem>
852 </varlistentry>
853
854 <varlistentry>
855 <term><option>-v4</option></term>
856 <listitem>
857 <para>The same as <option>-v3</option> except that the
858 intermediate program representation after each
859 compilation phase is also printed (excluding
860 preprocessed and C/assembly files).</para>
861 </listitem>
862 </varlistentry>
863 </variablelist>
864 </listitem>
865 </varlistentry>
866
867 <varlistentry>
868 <term><option>-ferror-spans</option>
869 <indexterm><primary><option>-ferror-spans</option></primary>
870 </indexterm>
871 </term>
872 <listitem>
873 <para>Causes GHC to emit the full source span of the
874 syntactic entity relating to an error message. Normally, GHC
875 emits the source location of the start of the syntactic
876 entity only.</para>
877
878 <para>For example:</para>
879
880 <screen>
881 test.hs:3:6: parse error on input `where'
882 </screen>
883
884 <para>becomes:</para>
885
886 <screen>
887 test296.hs:3:6-10: parse error on input `where'
888 </screen>
889
890 <para>And multi-line spans are possible too:</para>
891
892 <screen>
893 test.hs:(5,4)-(6,7):
894 Conflicting definitions for `a'
895 Bound at: test.hs:5:4
896 test.hs:6:7
897 In the binding group for: a, b, a
898 </screen>
899
900 <para>Note that line numbers start counting at one, but
901 column numbers start at zero. This choice was made to
902 follow existing convention (i.e. this is how Emacs does
903 it).</para>
904 </listitem>
905 </varlistentry>
906
907 <varlistentry>
908 <term><option>-H</option><replaceable>size</replaceable>
909 <indexterm><primary><option>-H</option></primary></indexterm>
910 </term>
911 <listitem>
912 <para>Set the minimum size of the heap to
913 <replaceable>size</replaceable>.
914 This option is equivalent to
915 <literal>+RTS&nbsp;-H<replaceable>size</replaceable></literal>,
916 see <xref linkend="rts-options-gc" />.
917 </para>
918 </listitem>
919 </varlistentry>
920
921 <varlistentry>
922 <term><option>-Rghc-timing</option>
923 <indexterm><primary><option>-Rghc-timing</option></primary></indexterm>
924 </term>
925 <listitem>
926 <para>Prints a one-line summary of timing statistics for the
927 GHC run. This option is equivalent to
928 <literal>+RTS&nbsp;-tstderr</literal>, see <xref
929 linkend="rts-options-gc" />.
930 </para>
931 </listitem>
932 </varlistentry>
933 </variablelist>
934 </sect1>
935
936 &separate;
937
938 <sect1 id="options-sanity">
939 <title>Warnings and sanity-checking</title>
940
941 <indexterm><primary>sanity-checking options</primary></indexterm>
942 <indexterm><primary>warnings</primary></indexterm>
943
944
945 <para>GHC has a number of options that select which types of
946 non-fatal error messages, otherwise known as warnings, can be
947 generated during compilation. By default, you get a standard set
948 of warnings which are generally likely to indicate bugs in your
949 program. These are:
950 <option>-fwarn-overlapping-patterns</option>,
951 <option>-fwarn-warnings-deprecations</option>,
952 <option>-fwarn-deprecated-flags</option>,
953 <option>-fwarn-duplicate-constraints</option>,
954 <option>-fwarn-duplicate-exports</option>,
955 <option>-fwarn-missing-fields</option>,
956 <option>-fwarn-missing-methods</option>,
957 <option>-fwarn-lazy-unlifted-bindings</option>,
958 <option>-fwarn-wrong-do-bind</option>,
959 <option>-fwarn-unsupported-calling-conventions</option>,
960 <option>-fwarn-dodgy-foreign-imports</option>, and
961 <option>-fwarn-typeable-instances</option>. The following
962 flags are
963 simple ways to select standard &ldquo;packages&rdquo; of warnings:
964 </para>
965
966 <variablelist>
967
968 <varlistentry>
969 <term><option>-W</option>:</term>
970 <listitem>
971 <indexterm><primary>-W option</primary></indexterm>
972 <para>Provides the standard warnings plus
973 <option>-fwarn-incomplete-patterns</option>,
974 <option>-fwarn-dodgy-exports</option>,
975 <option>-fwarn-dodgy-imports</option>,
976 <option>-fwarn-unused-matches</option>,
977 <option>-fwarn-unused-imports</option>, and
978 <option>-fwarn-unused-binds</option>.</para>
979 </listitem>
980 </varlistentry>
981
982 <varlistentry>
983 <term><option>-Wall</option>:</term>
984 <listitem>
985 <indexterm><primary><option>-Wall</option></primary></indexterm>
986 <para>Turns on all warning options that indicate potentially
987 suspicious code. The warnings that are
988 <emphasis>not</emphasis> enabled by <option>-Wall</option>
989 are
990 <option>-fwarn-tabs</option>,
991 <option>-fwarn-incomplete-uni-patterns</option>,
992 <option>-fwarn-incomplete-record-updates</option>,
993 <option>-fwarn-monomorphism-restriction</option>,
994 <option>-fwarn-auto-orphans</option>,
995 <option>-fwarn-implicit-prelude</option>,
996 <option>-fwarn-missing-local-sigs</option>,
997 <option>-fwarn-missing-import-lists</option>.</para>
998 </listitem>
999 </varlistentry>
1000
1001 <varlistentry>
1002 <term><option>-w</option>:</term>
1003 <listitem>
1004 <indexterm><primary><option>-w</option></primary></indexterm>
1005 <para>Turns off all warnings, including the standard ones and
1006 those that <literal>-Wall</literal> doesn't enable.</para>
1007 </listitem>
1008 </varlistentry>
1009
1010 <varlistentry>
1011 <term><option>-Werror</option>:</term>
1012 <listitem>
1013 <indexterm><primary><option>-Werror</option></primary></indexterm>
1014 <para>Makes any warning into a fatal error. Useful so that you don't
1015 miss warnings when doing batch compilation. </para>
1016 </listitem>
1017 </varlistentry>
1018
1019 <varlistentry>
1020 <term><option>-Wwarn</option>:</term>
1021 <listitem>
1022 <indexterm><primary><option>-Wwarn</option></primary></indexterm>
1023 <para>Warnings are treated only as warnings, not as errors. This is
1024 the default, but can be useful to negate a
1025 <option>-Werror</option> flag.</para>
1026 </listitem>
1027 </varlistentry>
1028
1029 </variablelist>
1030
1031 <para>The full set of warning options is described below. To turn
1032 off any warning, simply give the corresponding
1033 <option>-fno-warn-...</option> option on the command line.</para>
1034
1035 <variablelist>
1036
1037 <varlistentry>
1038 <term><option>-fdefer-type-errors</option>:</term>
1039 <listitem>
1040 <indexterm><primary><option>-fdefer-type-errors</option></primary>
1041 </indexterm>
1042 <indexterm><primary>warnings</primary></indexterm>
1043 <para>Defer as many type errors as possible until runtime.
1044 At compile time you get a warning (instead of an error). At
1045 runtime, if you use a value that depends on a type error, you
1046 get a runtime error; but you can run any type-correct parts of your code
1047 just fine. See <xref linkend="defer-type-errors"/></para>
1048 </listitem>
1049 </varlistentry>
1050
1051 <varlistentry>
1052 <term><option>-fhelpful-errors</option>:</term>
1053 <listitem>
1054 <indexterm><primary><option>-fhelpful-errors</option></primary>
1055 </indexterm>
1056 <indexterm><primary>warnings</primary></indexterm>
1057 <para>When a name or package is not found in scope, make
1058 suggestions for the name or package you might have meant instead.</para>
1059 <para>This option is on by default.</para>
1060 </listitem>
1061 </varlistentry>
1062
1063 <varlistentry>
1064 <term><option>-fwarn-unrecognised-pragmas</option>:</term>
1065 <listitem>
1066 <indexterm><primary><option>-fwarn-unrecognised-pragmas</option></primary>
1067 </indexterm>
1068 <indexterm><primary>warnings</primary></indexterm>
1069 <indexterm><primary>pragmas</primary></indexterm>
1070 <para>Causes a warning to be emitted when a
1071 pragma that GHC doesn't recognise is used. As well as pragmas
1072 that GHC itself uses, GHC also recognises pragmas known to be used
1073 by other tools, e.g. <literal>OPTIONS_HUGS</literal> and
1074 <literal>DERIVE</literal>.</para>
1075
1076 <para>This option is on by default.</para>
1077 </listitem>
1078 </varlistentry>
1079
1080 <varlistentry>
1081 <term><option>-fwarn-warnings-deprecations</option>:</term>
1082 <listitem>
1083 <indexterm><primary><option>-fwarn-warnings-deprecations</option></primary>
1084 </indexterm>
1085 <indexterm><primary>warnings</primary></indexterm>
1086 <indexterm><primary>deprecations</primary></indexterm>
1087 <para>Causes a warning to be emitted when a
1088 module, function or type with a WARNING or DEPRECATED pragma
1089 is used. See <xref linkend="warning-deprecated-pragma"/> for more
1090 details on the pragmas.</para>
1091
1092 <para>This option is on by default.</para>
1093 </listitem>
1094 </varlistentry>
1095
1096 <varlistentry>
1097 <term><option>-fwarn-deprecated-flags</option>:</term>
1098 <listitem>
1099 <indexterm><primary><option>-fwarn-deprecated-flags</option></primary>
1100 </indexterm>
1101 <indexterm><primary>deprecated-flags</primary></indexterm>
1102 <para>Causes a warning to be emitted when a deprecated
1103 commandline flag is used.</para>
1104
1105 <para>This option is on by default.</para>
1106 </listitem>
1107 </varlistentry>
1108
1109 <varlistentry>
1110 <term><option>-fwarn-unsupported-calling-conventions</option>:</term>
1111 <listitem>
1112 <indexterm><primary><option>-fwarn-unsupported-calling-conventions</option></primary>
1113 </indexterm>
1114 <para>Causes a warning to be emitted for foreign declarations
1115 that use unsupported calling conventions. In particular,
1116 if the <literal>stdcall</literal> calling convention is used
1117 on an architecture other than i386 then it will be treated
1118 as <literal>ccall</literal>.</para>
1119 </listitem>
1120 </varlistentry>
1121
1122 <varlistentry>
1123 <term><option>-fwarn-dodgy-foreign-imports</option>:</term>
1124 <listitem>
1125 <indexterm><primary><option>-fwarn-dodgy-foreign-imports</option></primary>
1126 </indexterm>
1127 <para>Causes a warning to be emitted for foreign imports of
1128 the following form:</para>
1129
1130 <programlisting>
1131 foreign import "f" f :: FunPtr t
1132 </programlisting>
1133
1134 <para>on the grounds that it probably should be</para>
1135
1136 <programlisting>
1137 foreign import "&amp;f" f :: FunPtr t
1138 </programlisting>
1139
1140 <para>The first form declares that `f` is a (pure) C
1141 function that takes no arguments and returns a pointer to a
1142 C function with type `t`, whereas the second form declares
1143 that `f` itself is a C function with type `t`. The first
1144 declaration is usually a mistake, and one that is hard to
1145 debug because it results in a crash, hence this
1146 warning.</para>
1147 </listitem>
1148 </varlistentry>
1149
1150 <varlistentry>
1151 <term><option>-fwarn-dodgy-exports</option>:</term>
1152 <listitem>
1153 <indexterm><primary><option>-fwarn-dodgy-exports</option></primary>
1154 </indexterm>
1155 <para>Causes a warning to be emitted when a datatype
1156 <literal>T</literal> is exported
1157 with all constructors, i.e. <literal>T(..)</literal>, but is it
1158 just a type synonym.</para>
1159 <para>Also causes a warning to be emitted when a module is
1160 re-exported, but that module exports nothing.</para>
1161 </listitem>
1162 </varlistentry>
1163
1164 <varlistentry>
1165 <term><option>-fwarn-dodgy-imports</option>:</term>
1166 <listitem>
1167 <indexterm><primary><option>-fwarn-dodgy-imports</option></primary>
1168 </indexterm>
1169 <para>Causes a warning to be emitted in the following cases:</para>
1170 <itemizedlist>
1171 <listitem>
1172 <para>When a datatype <literal>T</literal> is imported with all
1173 constructors, i.e. <literal>T(..)</literal>, but has been
1174 exported abstractly, i.e. <literal>T</literal>.
1175 </para>
1176 </listitem>
1177 <listitem>
1178 <para>When an <literal>import</literal> statement hides an
1179 entity that is not exported.</para>
1180 </listitem>
1181 </itemizedlist>
1182 </listitem>
1183 </varlistentry>
1184
1185 <varlistentry>
1186 <term><option>-fwarn-lazy-unlifted-bindings</option>:</term>
1187 <listitem>
1188 <indexterm><primary><option>-fwarn-lazy-unlifted-bindings</option></primary>
1189 </indexterm>
1190 <para>Causes a warning to be emitted when an unlifted type
1191 is bound in a way that looks lazy, e.g.
1192 <literal>where (I# x) = ...</literal>. Use
1193 <literal>where !(I# x) = ...</literal> instead. This will be an
1194 error, rather than a warning, in GHC 7.2.
1195 </para>
1196 </listitem>
1197 </varlistentry>
1198
1199 <varlistentry>
1200 <term><option>-fwarn-duplicate-constraints</option>:</term>
1201 <listitem>
1202 <indexterm><primary><option>-fwarn-duplicate-constraints</option></primary></indexterm>
1203 <indexterm><primary>duplicate constraints, warning</primary></indexterm>
1204
1205 <para>Have the compiler warn about duplicate constraints in a type signature. For
1206 example
1207 <programlisting>
1208 f :: (Eq a, Show a, Eq a) => a -> a
1209 </programlisting>
1210 The warning will indicate the duplicated <literal>Eq a</literal> constraint.
1211 </para>
1212
1213 <para>This option is on by default.</para>
1214 </listitem>
1215 </varlistentry>
1216
1217 <varlistentry>
1218 <term><option>-fwarn-duplicate-exports</option>:</term>
1219 <listitem>
1220 <indexterm><primary><option>-fwarn-duplicate-exports</option></primary></indexterm>
1221 <indexterm><primary>duplicate exports, warning</primary></indexterm>
1222 <indexterm><primary>export lists, duplicates</primary></indexterm>
1223
1224 <para>Have the compiler warn about duplicate entries in
1225 export lists. This is useful information if you maintain
1226 large export lists, and want to avoid the continued export
1227 of a definition after you've deleted (one) mention of it in
1228 the export list.</para>
1229
1230 <para>This option is on by default.</para>
1231 </listitem>
1232 </varlistentry>
1233
1234 <varlistentry>
1235 <term><option>-fwarn-hi-shadowing</option>:</term>
1236 <listitem>
1237 <indexterm><primary><option>-fwarn-hi-shadowing</option></primary></indexterm>
1238 <indexterm><primary>shadowing</primary>
1239 <secondary>interface files</secondary></indexterm>
1240
1241 <para>Causes the compiler to emit a warning when a module or
1242 interface file in the current directory is shadowing one
1243 with the same module name in a library or other
1244 directory.</para>
1245 </listitem>
1246 </varlistentry>
1247
1248 <varlistentry>
1249 <term><option>-fwarn-identities</option>:</term>
1250 <listitem>
1251 <indexterm><primary><option>-fwarn-identities</option></primary></indexterm>
1252 <para>Causes the compiler to emit a warning when a Prelude numeric
1253 conversion converts a type T to the same type T; such calls
1254 are probably no-ops and can be omitted. The functions checked for
1255 are: <literal>toInteger</literal>,
1256 <literal>toRational</literal>,
1257 <literal>fromIntegral</literal>,
1258 and <literal>realToFrac</literal>.
1259 </para>
1260 </listitem>
1261 </varlistentry>
1262
1263 <varlistentry>
1264 <term><option>-fwarn-implicit-prelude</option>:</term>
1265 <listitem>
1266 <indexterm><primary><option>-fwarn-implicit-prelude</option></primary></indexterm>
1267 <indexterm><primary>implicit prelude, warning</primary></indexterm>
1268 <para>Have the compiler warn if the Prelude is implicitly
1269 imported. This happens unless either the Prelude module is
1270 explicitly imported with an <literal>import ... Prelude ...</literal>
1271 line, or this implicit import is disabled (either by
1272 <option>-XNoImplicitPrelude</option> or a
1273 <literal>LANGUAGE NoImplicitPrelude</literal> pragma).</para>
1274
1275 <para>Note that no warning is given for syntax that implicitly
1276 refers to the Prelude, even if <option>-XNoImplicitPrelude</option>
1277 would change whether it refers to the Prelude.
1278 For example, no warning is given when
1279 <literal>368</literal> means
1280 <literal>Prelude.fromInteger (368::Prelude.Integer)</literal>
1281 (where <literal>Prelude</literal> refers to the actual Prelude module,
1282 regardless of the imports of the module being compiled).</para>
1283
1284 <para>This warning is off by default.</para>
1285 </listitem>
1286 </varlistentry>
1287
1288 <varlistentry>
1289 <term><option>-fwarn-incomplete-patterns</option>,
1290 <option>-fwarn-incomplete-uni-patterns</option>:
1291 </term>
1292 <listitem>
1293 <indexterm><primary><option>-fwarn-incomplete-patterns</option></primary></indexterm>
1294 <indexterm><primary><option>-fwarn-incomplete-uni-patterns</option></primary></indexterm>
1295 <indexterm><primary>incomplete patterns, warning</primary></indexterm>
1296 <indexterm><primary>patterns, incomplete</primary></indexterm>
1297
1298 <para>The option <option>-fwarn-incomplete-patterns</option> warns
1299 about places where
1300 a pattern-match might fail at runtime.
1301 The function
1302 <function>g</function> below will fail when applied to
1303 non-empty lists, so the compiler will emit a warning about
1304 this when <option>-fwarn-incomplete-patterns</option> is
1305 enabled.
1306
1307 <programlisting>
1308 g [] = 2
1309 </programlisting>
1310
1311 This option isn't enabled by default because it can be
1312 a bit noisy, and it doesn't always indicate a bug in the
1313 program. However, it's generally considered good practice
1314 to cover all the cases in your functions, and it is switched
1315 on by <option>-W</option>.</para>
1316
1317 <para>The flag <option>-fwarn-incomplete-uni-patterns</option> is
1318 similar, except that it
1319 applies only to lambda-expressions and pattern bindings, constructs
1320 that only allow a single pattern:
1321
1322 <programlisting>
1323 h = \[] -> 2
1324 Just k = f y
1325 </programlisting>
1326
1327 </para>
1328 </listitem>
1329 </varlistentry>
1330
1331 <varlistentry>
1332 <term><option>-fwarn-incomplete-record-updates</option>:</term>
1333 <listitem>
1334 <indexterm><primary><option>-fwarn-incomplete-record-updates</option></primary></indexterm>
1335 <indexterm><primary>incomplete record updates, warning</primary></indexterm>
1336 <indexterm><primary>record updates, incomplete</primary></indexterm>
1337
1338 <para>The function
1339 <function>f</function> below will fail when applied to
1340 <literal>Bar</literal>, so the compiler will emit a warning about
1341 this when <option>-fwarn-incomplete-record-updates</option> is
1342 enabled.</para>
1343
1344 <programlisting>
1345 data Foo = Foo { x :: Int }
1346 | Bar
1347
1348 f :: Foo -> Foo
1349 f foo = foo { x = 6 }
1350 </programlisting>
1351
1352 <para>This option isn't enabled by default because it can be
1353 very noisy, and it often doesn't indicate a bug in the
1354 program.</para>
1355 </listitem>
1356 </varlistentry>
1357
1358 <varlistentry>
1359 <term>
1360 <option>-fwarn-missing-fields</option>:
1361 <indexterm><primary><option>-fwarn-missing-fields</option></primary></indexterm>
1362 <indexterm><primary>missing fields, warning</primary></indexterm>
1363 <indexterm><primary>fields, missing</primary></indexterm>
1364 </term>
1365 <listitem>
1366
1367 <para>This option is on by default, and warns you whenever
1368 the construction of a labelled field constructor isn't
1369 complete, missing initializers for one or more fields. While
1370 not an error (the missing fields are initialised with
1371 bottoms), it is often an indication of a programmer error.</para>
1372 </listitem>
1373 </varlistentry>
1374
1375 <varlistentry>
1376 <term>
1377 <option>-fwarn-missing-import-lists</option>:
1378 <indexterm><primary><option>-fwarn-import-lists</option></primary></indexterm>
1379 <indexterm><primary>missing import lists, warning</primary></indexterm>
1380 <indexterm><primary>import lists, missing</primary></indexterm>
1381 </term>
1382 <listitem>
1383
1384 <para>This flag warns if you use an unqualified
1385 <literal>import</literal> declaration
1386 that does not explicitly list the entities brought into scope. For
1387 example
1388 </para>
1389
1390 <programlisting>
1391 module M where
1392 import X( f )
1393 import Y
1394 import qualified Z
1395 p x = f x x
1396 </programlisting>
1397
1398 <para>
1399 The <option>-fwarn-import-lists</option> flag will warn about the import
1400 of <literal>Y</literal> but not <literal>X</literal>
1401 If module <literal>Y</literal> is later changed to export (say) <literal>f</literal>,
1402 then the reference to <literal>f</literal> in <literal>M</literal> will become
1403 ambiguous. No warning is produced for the import of <literal>Z</literal>
1404 because extending <literal>Z</literal>'s exports would be unlikely to produce
1405 ambiguity in <literal>M</literal>.
1406 </para>
1407 </listitem>
1408 </varlistentry>
1409
1410 <varlistentry>
1411 <term><option>-fwarn-missing-methods</option>:</term>
1412 <listitem>
1413 <indexterm><primary><option>-fwarn-missing-methods</option></primary></indexterm>
1414 <indexterm><primary>missing methods, warning</primary></indexterm>
1415 <indexterm><primary>methods, missing</primary></indexterm>
1416
1417 <para>This option is on by default, and warns you whenever
1418 an instance declaration is missing one or more methods, and
1419 the corresponding class declaration has no default
1420 declaration for them.</para>
1421 <para>The warning is suppressed if the method name
1422 begins with an underscore. Here's an example where this is useful:
1423 <programlisting>
1424 class C a where
1425 _simpleFn :: a -> String
1426 complexFn :: a -> a -> String
1427 complexFn x y = ... _simpleFn ...
1428 </programlisting>
1429 The idea is that: (a) users of the class will only call <literal>complexFn</literal>;
1430 never <literal>_simpleFn</literal>; and (b)
1431 instance declarations can define either <literal>complexFn</literal> or <literal>_simpleFn</literal>.
1432 </para>
1433 </listitem>
1434 </varlistentry>
1435
1436 <varlistentry>
1437 <term><option>-fwarn-missing-signatures</option>:</term>
1438 <listitem>
1439 <indexterm><primary><option>-fwarn-missing-signatures</option></primary></indexterm>
1440 <indexterm><primary>type signatures, missing</primary></indexterm>
1441
1442 <para>If you would like GHC to check that every top-level
1443 function/value has a type signature, use the
1444 <option>-fwarn-missing-signatures</option> option. As part of
1445 the warning GHC also reports the inferred type. The
1446 option is off by default.</para>
1447 </listitem>
1448 </varlistentry>
1449
1450 <varlistentry>
1451 <term><option>-fwarn-missing-local-sigs</option>:</term>
1452 <listitem>
1453 <indexterm><primary><option>-fwarn-missing-local-sigs</option></primary></indexterm>
1454 <indexterm><primary>type signatures, missing</primary></indexterm>
1455
1456 <para>If you use the
1457 <option>-fwarn-missing-local-sigs</option> flag GHC will warn
1458 you about any polymorphic local bindings. As part of
1459 the warning GHC also reports the inferred type. The
1460 option is off by default.</para>
1461 </listitem>
1462 </varlistentry>
1463
1464 <varlistentry>
1465 <term><option>-fwarn-name-shadowing</option>:</term>
1466 <listitem>
1467 <indexterm><primary><option>-fwarn-name-shadowing</option></primary></indexterm>
1468 <indexterm><primary>shadowing, warning</primary></indexterm>
1469
1470 <para>This option causes a warning to be emitted whenever an
1471 inner-scope value has the same name as an outer-scope value,
1472 i.e. the inner value shadows the outer one. This can catch
1473 typographical errors that turn into hard-to-find bugs, e.g.,
1474 in the inadvertent capture of what would be a recursive call in
1475 <literal>f = ... let f = id in ... f ...</literal>.</para>
1476 <para>The warning is suppressed for names beginning with an underscore. For example
1477 <programlisting>
1478 f x = do { _ignore &lt;- this; _ignore &lt;- that; return (the other) }
1479 </programlisting>
1480 </para>
1481 </listitem>
1482 </varlistentry>
1483
1484 <varlistentry>
1485 <term><option>-fwarn-orphans, -fwarn-auto-orphans</option>:</term>
1486 <listitem>
1487 <indexterm><primary><option>-fwarn-orphans</option></primary></indexterm>
1488 <indexterm><primary><option>-fwarn-auto-orphans</option></primary></indexterm>
1489 <indexterm><primary>orphan instances, warning</primary></indexterm>
1490 <indexterm><primary>orphan rules, warning</primary></indexterm>
1491
1492 <para>These flags cause a warning to be emitted whenever the
1493 module contains an "orphan" instance declaration or rewrite rule.
1494 An instance declaration is an orphan if it appears in a module in
1495 which neither the class nor the type being instanced are declared
1496 in the same module. A rule is an orphan if it is a rule for a
1497 function declared in another module. A module containing any
1498 orphans is called an orphan module.</para>
1499 <para>The trouble with orphans is that GHC must pro-actively read the interface
1500 files for all orphan modules, just in case their instances or rules
1501 play a role, whether or not the module's interface would otherwise
1502 be of any use. See <xref linkend="orphan-modules"/> for details.
1503 </para>
1504 <para>The flag <option>-fwarn-orphans</option> warns about user-written
1505 orphan rules or instances. The flag <option>-fwarn-auto-orphans</option>
1506 warns about automatically-generated orphan rules, notably as a result of
1507 specialising functions, for type classes (<literal>Specialise</literal>)
1508 or argument values (<literal>SpecConstr</literal>).</para>
1509 </listitem>
1510 </varlistentry>
1511
1512 <varlistentry>
1513 <term>
1514 <option>-fwarn-overlapping-patterns</option>:
1515 <indexterm><primary><option>-fwarn-overlapping-patterns</option></primary></indexterm>
1516 <indexterm><primary>overlapping patterns, warning</primary></indexterm>
1517 <indexterm><primary>patterns, overlapping</primary></indexterm>
1518 </term>
1519 <listitem>
1520 <para>By default, the compiler will warn you if a set of
1521 patterns are overlapping, e.g.,</para>
1522
1523 <programlisting>
1524 f :: String -&#62; Int
1525 f [] = 0
1526 f (_:xs) = 1
1527 f "2" = 2
1528 </programlisting>
1529
1530 <para>where the last pattern match in <function>f</function>
1531 won't ever be reached, as the second pattern overlaps
1532 it. More often than not, redundant patterns is a programmer
1533 mistake/error, so this option is enabled by default.</para>
1534 </listitem>
1535 </varlistentry>
1536
1537 <varlistentry>
1538 <term><option>-fwarn-tabs</option>:</term>
1539 <listitem>
1540 <indexterm><primary><option>-fwarn-tabs</option></primary></indexterm>
1541 <indexterm><primary>tabs, warning</primary></indexterm>
1542 <para>Have the compiler warn if there are tabs in your source
1543 file.</para>
1544
1545 <para>This warning is off by default.</para>
1546 </listitem>
1547 </varlistentry>
1548
1549 <varlistentry>
1550 <term><option>-fwarn-type-defaults</option>:</term>
1551 <listitem>
1552 <indexterm><primary><option>-fwarn-type-defaults</option></primary></indexterm>
1553 <indexterm><primary>defaulting mechanism, warning</primary></indexterm>
1554 <para>Have the compiler warn/inform you where in your source
1555 the Haskell defaulting mechanism for numeric types kicks
1556 in. This is useful information when converting code from a
1557 context that assumed one default into one with another,
1558 e.g., the &lsquo;default default&rsquo; for Haskell 1.4 caused the
1559 otherwise unconstrained value <constant>1</constant> to be
1560 given the type <literal>Int</literal>, whereas Haskell 98
1561 and later
1562 defaults it to <literal>Integer</literal>. This may lead to
1563 differences in performance and behaviour, hence the
1564 usefulness of being non-silent about this.</para>
1565
1566 <para>This warning is off by default.</para>
1567 </listitem>
1568 </varlistentry>
1569
1570 <varlistentry>
1571 <term><option>-fwarn-monomorphism-restriction</option>:</term>
1572 <listitem>
1573 <indexterm><primary><option>-fwarn-monomorphism-restriction</option></primary></indexterm>
1574 <indexterm><primary>monomorphism restriction, warning</primary></indexterm>
1575 <para>Have the compiler warn/inform you where in your source
1576 the Haskell Monomorphism Restriction is applied. If applied silently
1577 the MR can give rise to unexpected behaviour, so it can be helpful
1578 to have an explicit warning that it is being applied.</para>
1579
1580 <para>This warning is off by default.</para>
1581 </listitem>
1582 </varlistentry>
1583
1584 <varlistentry>
1585 <term><option>-fwarn-unused-binds</option>:</term>
1586 <listitem>
1587 <indexterm><primary><option>-fwarn-unused-binds</option></primary></indexterm>
1588 <indexterm><primary>unused binds, warning</primary></indexterm>
1589 <indexterm><primary>binds, unused</primary></indexterm>
1590 <para>Report any function definitions (and local bindings)
1591 which are unused. For top-level functions, the warning is
1592 only given if the binding is not exported.</para>
1593 <para>A definition is regarded as "used" if (a) it is exported, or (b) it is
1594 mentioned in the right hand side of another definition that is used, or (c) the
1595 function it defines begins with an underscore. The last case provides a
1596 way to suppress unused-binding warnings selectively. </para>
1597 <para> Notice that a variable
1598 is reported as unused even if it appears in the right-hand side of another
1599 unused binding. </para>
1600 </listitem>
1601 </varlistentry>
1602
1603 <varlistentry>
1604 <term><option>-fwarn-unused-imports</option>:</term>
1605 <listitem>
1606 <indexterm><primary><option>-fwarn-unused-imports</option></primary></indexterm>
1607 <indexterm><primary>unused imports, warning</primary></indexterm>
1608 <indexterm><primary>imports, unused</primary></indexterm>
1609
1610 <para>Report any modules that are explicitly imported but
1611 never used. However, the form <literal>import M()</literal> is
1612 never reported as an unused import, because it is a useful idiom
1613 for importing instance declarations, which are anonymous in Haskell.</para>
1614 </listitem>
1615 </varlistentry>
1616
1617 <varlistentry>
1618 <term><option>-fwarn-unused-matches</option>:</term>
1619 <listitem>
1620 <indexterm><primary><option>-fwarn-unused-matches</option></primary></indexterm>
1621 <indexterm><primary>unused matches, warning</primary></indexterm>
1622 <indexterm><primary>matches, unused</primary></indexterm>
1623
1624 <para>Report all unused variables which arise from pattern
1625 matches, including patterns consisting of a single variable.
1626 For instance <literal>f x y = []</literal> would report
1627 <varname>x</varname> and <varname>y</varname> as unused. The
1628 warning is suppressed if the variable name begins with an underscore, thus:
1629 <programlisting>
1630 f _x = True
1631 </programlisting>
1632 </para>
1633 </listitem>
1634 </varlistentry>
1635
1636 <varlistentry>
1637 <term><option>-fwarn-unused-do-bind</option>:</term>
1638 <listitem>
1639 <indexterm><primary><option>-fwarn-unused-do-bind</option></primary></indexterm>
1640 <indexterm><primary>unused do binding, warning</primary></indexterm>
1641 <indexterm><primary>do binding, unused</primary></indexterm>
1642
1643 <para>Report expressions occurring in <literal>do</literal> and <literal>mdo</literal> blocks
1644 that appear to silently throw information away.
1645 For instance <literal>do { mapM popInt xs ; return 10 }</literal> would report
1646 the first statement in the <literal>do</literal> block as suspicious,
1647 as it has the type <literal>StackM [Int]</literal> and not <literal>StackM ()</literal>, but that
1648 <literal>[Int]</literal> value is not bound to anything. The warning is suppressed by
1649 explicitly mentioning in the source code that your program is throwing something away:
1650 <programlisting>
1651 do { _ &lt;- mapM popInt xs ; return 10 }
1652 </programlisting>
1653 Of course, in this particular situation you can do even better:
1654 <programlisting>
1655 do { mapM_ popInt xs ; return 10 }
1656 </programlisting>
1657 </para>
1658 </listitem>
1659 </varlistentry>
1660
1661 <varlistentry>
1662 <term><option>-fwarn-wrong-do-bind</option>:</term>
1663 <listitem>
1664 <indexterm><primary><option>-fwarn-wrong-do-bind</option></primary></indexterm>
1665 <indexterm><primary>apparently erroneous do binding, warning</primary></indexterm>
1666 <indexterm><primary>do binding, apparently erroneous</primary></indexterm>
1667
1668 <para>Report expressions occurring in <literal>do</literal> and <literal>mdo</literal> blocks
1669 that appear to lack a binding.
1670 For instance <literal>do { return (popInt 10) ; return 10 }</literal> would report
1671 the first statement in the <literal>do</literal> block as suspicious,
1672 as it has the type <literal>StackM (StackM Int)</literal> (which consists of two nested applications
1673 of the same monad constructor), but which is not then &quot;unpacked&quot; by binding the result.
1674 The warning is suppressed by explicitly mentioning in the source code that your program is throwing something away:
1675 <programlisting>
1676 do { _ &lt;- return (popInt 10) ; return 10 }
1677 </programlisting>
1678 For almost all sensible programs this will indicate a bug, and you probably intended to write:
1679 <programlisting>
1680 do { popInt 10 ; return 10 }
1681 </programlisting>
1682 </para>
1683 </listitem>
1684 </varlistentry>
1685
1686 <varlistentry>
1687 <term><option>-fwarn-typeable-instances</option>:</term>
1688 <listitem>
1689 <indexterm><primary><option>-fwarn-typeable-instances</option></primary></indexterm>
1690 <indexterm><primary>typeable instances, warning</primary></indexterm>
1691
1692 <para>Report handwritten (ie. not derived) instances of the
1693 <literal>Typeable</literal> class. These are ignore by the compiler;
1694 only derived instances of <literal>Typeable</literal> are allowed.
1695 </para>
1696 </listitem>
1697 </varlistentry>
1698
1699 </variablelist>
1700
1701 <para>If you're feeling really paranoid, the
1702 <option>-dcore-lint</option>
1703 option<indexterm><primary><option>-dcore-lint</option></primary></indexterm>
1704 is a good choice. It turns on heavyweight intra-pass
1705 sanity-checking within GHC. (It checks GHC's sanity, not
1706 yours.)</para>
1707
1708 </sect1>
1709
1710 &packages;
1711
1712 <sect1 id="options-optimise">
1713 <title>Optimisation (code improvement)</title>
1714
1715 <indexterm><primary>optimisation</primary></indexterm>
1716 <indexterm><primary>improvement, code</primary></indexterm>
1717
1718 <para>The <option>-O*</option> options specify convenient
1719 &ldquo;packages&rdquo; of optimisation flags; the
1720 <option>-f*</option> options described later on specify
1721 <emphasis>individual</emphasis> optimisations to be turned on/off;
1722 the <option>-m*</option> options specify
1723 <emphasis>machine-specific</emphasis> optimisations to be turned
1724 on/off.</para>
1725
1726 <sect2 id="optimise-pkgs">
1727 <title><option>-O*</option>: convenient &ldquo;packages&rdquo; of optimisation flags.</title>
1728
1729 <para>There are <emphasis>many</emphasis> options that affect
1730 the quality of code produced by GHC. Most people only have a
1731 general goal, something like &ldquo;Compile quickly&rdquo; or
1732 &ldquo;Make my program run like greased lightning.&rdquo; The
1733 following &ldquo;packages&rdquo; of optimisations (or lack
1734 thereof) should suffice.</para>
1735
1736 <para>Note that higher optimisation levels cause more
1737 cross-module optimisation to be performed, which can have an
1738 impact on how much of your program needs to be recompiled when
1739 you change something. This is one reason to stick to
1740 no-optimisation when developing code.</para>
1741
1742 <variablelist>
1743
1744 <varlistentry>
1745 <term>
1746 No <option>-O*</option>-type option specified:
1747 <indexterm><primary>-O* not specified</primary></indexterm>
1748 </term>
1749 <listitem>
1750 <para>This is taken to mean: &ldquo;Please compile
1751 quickly; I'm not over-bothered about compiled-code
1752 quality.&rdquo; So, for example: <command>ghc -c
1753 Foo.hs</command></para>
1754 </listitem>
1755 </varlistentry>
1756
1757 <varlistentry>
1758 <term>
1759 <option>-O0</option>:
1760 <indexterm><primary><option>-O0</option></primary></indexterm>
1761 </term>
1762 <listitem>
1763 <para>Means &ldquo;turn off all optimisation&rdquo;,
1764 reverting to the same settings as if no
1765 <option>-O</option> options had been specified. Saying
1766 <option>-O0</option> can be useful if
1767 eg. <command>make</command> has inserted a
1768 <option>-O</option> on the command line already.</para>
1769 </listitem>
1770 </varlistentry>
1771
1772 <varlistentry>
1773 <term>
1774 <option>-O</option> or <option>-O1</option>:
1775 <indexterm><primary>-O option</primary></indexterm>
1776 <indexterm><primary>-O1 option</primary></indexterm>
1777 <indexterm><primary>optimise</primary><secondary>normally</secondary></indexterm>
1778 </term>
1779 <listitem>
1780 <para>Means: &ldquo;Generate good-quality code without
1781 taking too long about it.&rdquo; Thus, for example:
1782 <command>ghc -c -O Main.lhs</command></para>
1783 </listitem>
1784 </varlistentry>
1785
1786 <varlistentry>
1787 <term>
1788 <option>-O2</option>:
1789 <indexterm><primary>-O2 option</primary></indexterm>
1790 <indexterm><primary>optimise</primary><secondary>aggressively</secondary></indexterm>
1791 </term>
1792 <listitem>
1793 <para>Means: &ldquo;Apply every non-dangerous
1794 optimisation, even if it means significantly longer
1795 compile times.&rdquo;</para>
1796
1797 <para>The avoided &ldquo;dangerous&rdquo; optimisations
1798 are those that can make runtime or space
1799 <emphasis>worse</emphasis> if you're unlucky. They are
1800 normally turned on or off individually.</para>
1801
1802 <para>At the moment, <option>-O2</option> is
1803 <emphasis>unlikely</emphasis> to produce better code than
1804 <option>-O</option>.</para>
1805 </listitem>
1806 </varlistentry>
1807 </variablelist>
1808
1809 <para>We don't use a <option>-O*</option> flag for day-to-day
1810 work. We use <option>-O</option> to get respectable speed;
1811 e.g., when we want to measure something. When we want to go for
1812 broke, we tend to use <option>-O2</option> (and we go for
1813 lots of coffee breaks).</para>
1814
1815 <para>The easiest way to see what <option>-O</option> (etc.)
1816 &ldquo;really mean&rdquo; is to run with <option>-v</option>,
1817 then stand back in amazement.</para>
1818 </sect2>
1819
1820 <sect2 id="options-f">
1821 <title><option>-f*</option>: platform-independent flags</title>
1822
1823 <indexterm><primary>-f* options (GHC)</primary></indexterm>
1824 <indexterm><primary>-fno-* options (GHC)</primary></indexterm>
1825
1826 <para>These flags turn on and off individual optimisations.
1827 They are normally set via the <option>-O</option> options
1828 described above, and as such, you shouldn't need to set any of
1829 them explicitly (indeed, doing so could lead to unexpected
1830 results). A flag <option>-fwombat</option> can be negated by
1831 saying <option>-fno-wombat</option>. The flags below are off
1832 by default, except where noted below.
1833 </para>
1834
1835 <variablelist>
1836 <varlistentry>
1837 <term>
1838 <option>-fcse</option>
1839 <indexterm><primary><option>-fcse</option></primary></indexterm>
1840 </term>
1841 <listitem>
1842 <para><emphasis>On by default.</emphasis>. Enables the common-sub-expression
1843 elimination optimisation.
1844 Switching this off can be useful if you have some <literal>unsafePerformIO</literal>
1845 expressions that you don't want commoned-up.</para>
1846 </listitem>
1847 </varlistentry>
1848
1849 <varlistentry>
1850 <term>
1851 <option>-fstrictness</option>
1852 <indexterm><primary><option></option></primary></indexterm>
1853 </term>
1854 <listitem>
1855 <para> <emphasis>On by default.</emphasis>.
1856 Switch on the strictness analyser. There is a very old paper about GHC's
1857 strictness analyser, <ulink url="http://research.microsoft.com/en-us/um/people/simonpj/papers/simple-strictnes-analyser.ps.gz">
1858 Measuring the effectiveness of a simple strictness analyser</ulink>,
1859 but the current one is quite a bit different.
1860 </para>
1861
1862 <para>The strictness analyser figures out when arguments and
1863 variables in a function can be treated 'strictly' (that is they
1864 are always evaluated in the function at some point). This allow
1865 GHC to apply certain optimisations such as unboxing that
1866 otherwise don't apply as they change the semantics of the program
1867 when applied to lazy arguments.
1868 </para>
1869 </listitem>
1870 </varlistentry>
1871
1872 <varlistentry>
1873 <term>
1874 <option>-funbox-strict-fields</option>:
1875 <indexterm><primary><option>-funbox-strict-fields</option></primary></indexterm>
1876 <indexterm><primary>strict constructor fields</primary></indexterm>
1877 <indexterm><primary>constructor fields, strict</primary></indexterm>
1878 </term>
1879 <listitem>
1880 <para>This option causes all constructor fields which are marked
1881 strict (i.e. &ldquo;!&rdquo;) to be unpacked if possible. It is
1882 equivalent to adding an <literal>UNPACK</literal> pragma to every
1883 strict constructor field (see <xref linkend="unpack-pragma"/>).
1884 </para>
1885
1886 <para>This option is a bit of a sledgehammer: it might sometimes
1887 make things worse. Selectively unboxing fields by using
1888 <literal>UNPACK</literal> pragmas might be better. An alternative
1889 is to use <option>-funbox-strict-fields</option> to turn on
1890 unboxing by default but disable it for certain constructor
1891 fields using the <literal>NOUNPACK</literal> pragma (see
1892 <xref linkend="nounpack-pragma"/>).</para>
1893 </listitem>
1894 </varlistentry>
1895
1896 <varlistentry>
1897 <term>
1898 <option>-funbox-small-strict-fields</option>:
1899 <indexterm><primary><option>-funbox-small-strict-fields</option></primary></indexterm>
1900 <indexterm><primary>strict constructor fields</primary></indexterm>
1901 <indexterm><primary>constructor fields, strict</primary></indexterm>
1902 </term>
1903 <listitem>
1904 <para>This option causes all constructor fields which are
1905 marked strict (i.e. &ldquo;!&rdquo;) and which
1906 representation is smaller or equal to the size of a
1907 pointer to be unpacked, if possible. It is equivalent to
1908 adding an <literal>UNPACK</literal> pragma (see <xref
1909 linkend="unpack-pragma"/>) to every strict constructor
1910 field that fulfils the size restriction.
1911 </para>
1912
1913 <para>For example, the constructor fields in the following
1914 data types
1915 <programlisting>
1916 data A = A !Int
1917 data B = B !A
1918 newtype C = C B
1919 data D = D !C
1920 </programlisting>
1921 would all be represented by a single
1922 <literal>Int#</literal> (see <xref linkend="primitives"/>)
1923 value with
1924 <option>-funbox-small-strict-fields</option> enabled.
1925 </para>
1926
1927 <para>This option is less of a sledgehammer than
1928 <option>-funbox-strict-fields</option>: it should rarely make things
1929 worse. If you use <option>-funbox-small-strict-fields</option>
1930 to turn on unboxing by default you can disable it for certain
1931 constructor fields using the <literal>NOUNPACK</literal> pragma (see
1932 <xref linkend="nounpack-pragma"/>).</para>
1933
1934 <para>
1935 Note that for consistency <literal>Double</literal>,
1936 <literal>Word64</literal>, and <literal>Int64</literal> constructor
1937 fields are unpacked on 32-bit platforms, even though they are
1938 technically larger than a pointer on those platforms.
1939 </para>
1940 </listitem>
1941 </varlistentry>
1942
1943 <varlistentry>
1944 <term>
1945 <option>-fspec-constr</option>
1946 <indexterm><primary><option>-fspec-constr</option></primary></indexterm>
1947 </term>
1948 <listitem>
1949 <para><emphasis>Off by default, but enabled by -O2.</emphasis>
1950 Turn on call-pattern specialisation; see
1951 <ulink url="http://research.microsoft.com/en-us/um/people/simonpj/papers/spec-constr/index.htm">
1952 Call-pattern specialisation for Haskell programs</ulink>.
1953 </para>
1954
1955 <para>This optimisation specializes recursive functions according to
1956 their argument "shapes". This is best explained by example so
1957 consider:
1958 <programlisting>
1959 last :: [a] -> a
1960 last [] = error "last"
1961 last (x : []) = x
1962 last (x : xs) = last xs
1963 </programlisting>
1964 In this code, once we pass the initial check for an empty list we
1965 know that in the recursive case this pattern match is redundant. As
1966 such <option>-fspec-constr</option> will transform the above code
1967 to:
1968 <programlisting>
1969 last :: [a] -> a
1970 last [] = error "last"
1971 last (x : xs) = last' x xs
1972 where
1973 last' x [] = x
1974 last' x (y : ys) = last' y ys
1975 </programlisting>
1976 </para>
1977
1978 <para>As well avoid unnecessary pattern matching it also helps avoid
1979 unnecessary allocation. This applies when a argument is strict in
1980 the recursive call to itself but not on the initial entry. As
1981 strict recursive branch of the function is created similar to the
1982 above example.
1983 </para>
1984 </listitem>
1985 </varlistentry>
1986
1987 <varlistentry>
1988 <term>
1989 <option>-fspecialise</option>
1990 <indexterm><primary><option>-fspecialise</option></primary></indexterm>
1991 </term>
1992 <listitem>
1993 <para><emphasis>On by default.</emphasis>
1994 Specialise each type-class-overloaded function defined in this
1995 module for the types at which it is called in this module. Also
1996 specialise imported functions that have an INLINABLE pragma
1997 (<xref linkend="inlinable-pragma"/>) for the types at which they
1998 are called in this module.
1999 </para>
2000 </listitem>
2001 </varlistentry>
2002
2003 <varlistentry>
2004 <term>
2005 <option>-fstatic-argument-transformation</option>
2006 <indexterm><primary><option>-fstatic-argument-transformation</option></primary></indexterm>
2007 </term>
2008 <listitem>
2009 <para>Turn on the static argument transformation, which turns a
2010 recursive function into a non-recursive one with a local
2011 recursive loop. See Chapter 7 of
2012 <ulink url="http://research.microsoft.com/en-us/um/people/simonpj/papers/santos-thesis.ps.gz">
2013 Andre Santos's PhD thesis</ulink>
2014 </para>
2015 </listitem>
2016 </varlistentry>
2017
2018 <varlistentry>
2019 <term>
2020 <option>-ffloat-in</option>
2021 <indexterm><primary><option></option></primary></indexterm>
2022 </term>
2023 <listitem>
2024 <para><emphasis>On by default.</emphasis>
2025 Float let-bindings inwards, nearer their binding site. See
2026 <ulink url="http://research.microsoft.com/en-us/um/people/simonpj/papers/float.ps.gz">
2027 Let-floating: moving bindings to give faster programs (ICFP'96)</ulink>.
2028 </para>
2029
2030 <para>This optimisation moves let bindings closer to their use
2031 site. The benefit here is that this may avoid unnecessary
2032 allocation if the branch the let is now on is never executed. It
2033 also enables other optimisation passes to work more effectively
2034 as they have more information locally.
2035 </para>
2036
2037 <para>This optimisation isn't always beneficial though (so GHC
2038 applies some heuristics to decide when to apply it). The details
2039 get complicated but a simple example is that it is often beneficial
2040 to move let bindings outwards so that multiple let bindings can be
2041 grouped into a larger single let binding, effectively batching
2042 their allocation and helping the garbage collector and allocator.
2043 </para>
2044 </listitem>
2045 </varlistentry>
2046
2047 <varlistentry>
2048 <term>
2049 <option>-ffull-laziness</option>
2050 <indexterm><primary><option>-ffull-laziness</option></primary></indexterm>
2051 </term>
2052 <listitem>
2053 <para><emphasis>On by default.</emphasis>
2054 Run the full laziness optimisation (also known as let-floating),
2055 which floats let-bindings outside enclosing lambdas, in the hope
2056 they will be thereby be computed less often. See
2057 <ulink url="http://research.microsoft.com/en-us/um/people/simonpj/papers/float.ps.gz">Let-floating:
2058 moving bindings to give faster programs (ICFP'96)</ulink>.
2059 Full laziness increases sharing, which can lead to increased memory
2060 residency.
2061 </para>
2062
2063 <para>NOTE: GHC doesn't implement complete full-laziness.
2064 When optimisation in on, and <option>-fno-full-laziness</option>
2065 is not given, some transformations that increase sharing are
2066 performed, such as extracting repeated computations from a loop.
2067 These are the same transformations that a fully lazy
2068 implementation would do, the difference is that GHC doesn't
2069 consistently apply full-laziness, so don't rely on it.
2070 </para>
2071 </listitem>
2072 </varlistentry>
2073
2074 <varlistentry>
2075 <term>
2076 <option>-fdo-lambda-eta-expansion</option>
2077 <indexterm><primary><option></option></primary></indexterm>
2078 </term>
2079 <listitem>
2080 <para><emphasis>On by default.</emphasis>
2081 Eta-expand let-bindings to increase their arity.
2082 </para>
2083 </listitem>
2084 </varlistentry>
2085
2086 <varlistentry>
2087 <term>
2088 <option>-fdo-eta-reduction</option>
2089 <indexterm><primary><option></option></primary></indexterm>
2090 </term>
2091 <listitem>
2092 <para><emphasis>On by default.</emphasis>
2093 Eta-reduce lambda expressions, if doing so gets rid of a whole
2094 group of lambdas.
2095 </para>
2096 </listitem>
2097 </varlistentry>
2098
2099 <varlistentry>
2100 <term>
2101 <option>-fcase-merge</option>
2102 <indexterm><primary><option></option></primary></indexterm>
2103 </term>
2104 <listitem>
2105 <para><emphasis>On by default.</emphasis>
2106 Merge immediately-nested case expressions that scrutinse the same variable. Example
2107 <programlisting>
2108 case x of
2109 Red -> e1
2110 _ -> case x of
2111 Blue -> e2
2112 Green -> e3
2113 ==>
2114 case x of
2115 Red -> e1
2116 Blue -> e2
2117 Green -> e2
2118 </programlisting>
2119 </para>
2120 </listitem>
2121 </varlistentry>
2122
2123 <varlistentry>
2124 <term>
2125 <option>-fliberate-case</option>
2126 <indexterm><primary><option>-fliberate-case</option></primary></indexterm>
2127 </term>
2128 <listitem>
2129 <para><emphasis>Off by default, but enabled by -O2.</emphasis>
2130 Turn on the liberate-case transformation. This unrolls recursive
2131 function once in its own RHS, to avoid repeated case analysis of
2132 free variables. It's a bit like the call-pattern specialiser
2133 (<option>-fspec-constr</option>) but for free variables rather than
2134 arguments.
2135 </para>
2136 </listitem>
2137 </varlistentry>
2138
2139 <varlistentry>
2140 <term>
2141 <option>-fdicts-cheap</option>
2142 <indexterm><primary><option></option></primary></indexterm>
2143 </term>
2144 <listitem>
2145 <para>A very experimental flag that makes dictionary-valued
2146 expressions seem cheap to the optimiser.
2147 </para>
2148 </listitem>
2149 </varlistentry>
2150
2151 <varlistentry>
2152 <term>
2153 <option>-feager-blackholing</option>
2154 <indexterm><primary><option></option></primary></indexterm>
2155 </term>
2156 <listitem>
2157 <para>Usually GHC black-holes a thunk only when it switches
2158 threads. This flag makes it do so as soon as the thunk is
2159 entered. See <ulink url="http://research.microsoft.com/en-us/um/people/simonpj/papers/parallel/">
2160 Haskell on a shared-memory multiprocessor</ulink>.
2161 </para>
2162 </listitem>
2163 </varlistentry>
2164
2165 <varlistentry>
2166 <term>
2167 <option>-fno-state-hack</option>
2168 <indexterm><primary><option>-fno-state-hack</option></primary></indexterm>
2169 </term>
2170 <listitem>
2171 <para>Turn off the "state hack" whereby any lambda with a
2172 <literal>State#</literal> token as argument is considered to be
2173 single-entry, hence it is considered OK to inline things inside
2174 it. This can improve performance of IO and ST monad code, but it
2175 runs the risk of reducing sharing.
2176 </para>
2177 </listitem>
2178 </varlistentry>
2179
2180 <varlistentry>
2181 <term>
2182 <option>-fpedantic-bottoms</option>
2183 <indexterm><primary><option>-fpedantic-bottoms</option></primary></indexterm>
2184 </term>
2185 <listitem>
2186 <para>Make GHC be more precise about its treatment of bottom (but see also
2187 <option>-fno-state-hack</option>). In particular, stop GHC
2188 eta-expanding through a case expression, which is good for
2189 performance, but bad if you are using <literal>seq</literal> on
2190 partial applications.
2191 </para>
2192 </listitem>
2193 </varlistentry>
2194
2195 <varlistentry>
2196 <term>
2197 <option>-fsimpl-tick-factor=<replaceable>n</replaceable></option>
2198 <indexterm><primary><option>-fsimpl-tick-factor</option></primary></indexterm>
2199 </term>
2200 <listitem>
2201 <para>GHC's optimiser can diverge if you write rewrite rules (
2202 <xref linkend="rewrite-rules"/>) that don't terminate, or (less
2203 satisfactorily) if you code up recursion through data types
2204 (<xref linkend="bugs-ghc"/>). To avoid making the compiler fall
2205 into an infinite loop, the optimiser carries a "tick count" and
2206 stops inlining and applying rewrite rules when this count is
2207 exceeded. The limit is set as a multiple of the program size, so
2208 bigger programs get more ticks. The
2209 <option>-fsimpl-tick-factor</option> flag lets you change the
2210 multiplier. The default is 100; numbers larger than 100 give more
2211 ticks, and numbers smaller than 100 give fewer.
2212 </para>
2213
2214 <para>If the tick-count expires, GHC summarises what simplifier
2215 steps it has done; you can use
2216 <option>-fddump-simpl-stats</option> to generate a much more
2217 detailed list. Usually that identifies the loop quite
2218 accurately, because some numbers are very large.
2219 </para>
2220 </listitem>
2221 </varlistentry>
2222
2223 <varlistentry>
2224 <term>
2225 <option>-funfolding-creation-threshold=<replaceable>n</replaceable></option>:
2226 <indexterm><primary><option>-funfolding-creation-threshold</option></primary></indexterm>
2227 <indexterm><primary>inlining, controlling</primary></indexterm>
2228 <indexterm><primary>unfolding, controlling</primary></indexterm>
2229 </term>
2230 <listitem>
2231 <para>(Default: 45) Governs the maximum size that GHC will allow a
2232 function unfolding to be. (An unfolding has a &ldquo;size&rdquo;
2233 that reflects the cost in terms of &ldquo;code bloat&rdquo; of
2234 expanding (aka inlining) that unfolding at a call site. A bigger
2235 function would be assigned a bigger cost.)
2236 </para>
2237
2238 <para>Consequences: (a) nothing larger than this will be inlined
2239 (unless it has an INLINE pragma); (b) nothing larger than this
2240 will be spewed into an interface file.
2241 </para>
2242
2243 <para>Increasing this figure is more likely to result in longer
2244 compile times than faster code. The
2245 <option>-funfolding-use-threshold</option> is more useful.
2246 </para>
2247 </listitem>
2248 </varlistentry>
2249
2250 <varlistentry>
2251 <term>
2252 <option>-funfolding-use-threshold=<replaceable>n</replaceable></option>
2253 <indexterm><primary><option>-funfolding-use-threshold</option></primary></indexterm>
2254 <indexterm><primary>inlining, controlling</primary></indexterm>
2255 <indexterm><primary>unfolding, controlling</primary></indexterm>
2256 </term>
2257 <listitem>
2258 <para>(Default: 8) This is the magic cut-off figure for unfolding
2259 (aka inlining): below this size, a function definition will be
2260 unfolded at the call-site, any bigger and it won't. The size
2261 computed for a function depends on two things: the actual size of
2262 the expression minus any discounts that
2263 apply (see <option>-funfolding-con-discount</option>).
2264 </para>
2265
2266 <para>The difference between this and
2267 <option>-funfolding-creation-threshold</option> is that this one
2268 determines if a function definition will be inlined <emphasis>at
2269 a call site</emphasis>. The other option determines if a
2270 function definition will be kept around at all for potential
2271 inlining.
2272 </para>
2273 </listitem>
2274 </varlistentry>
2275
2276 <varlistentry>
2277 <term>
2278 <option>-fexpose-all-unfoldings</option>
2279 <indexterm><primary><option></option></primary></indexterm>
2280 </term>
2281 <listitem>
2282 <para>An experimental flag to expose all unfoldings, even for very
2283 large or recursive functions. This allows for all functions to be
2284 inlined while usually GHC would avoid inlining larger functions.
2285 </para>
2286 </listitem>
2287 </varlistentry>
2288
2289 <varlistentry>
2290 <term>
2291 <option>-fvectorise</option>
2292 <indexterm><primary><option></option></primary></indexterm>
2293 </term>
2294 <listitem>
2295 <para>Part of <link linkend="dph">Data Parallel Haskell
2296 (DPH)</link>.</para>
2297
2298 <para><emphasis>Off by default.</emphasis> Enable the
2299 <emphasis>vectorisation</emphasis> optimisation transformation. This
2300 optimisation transforms the nested data parallelism code of programs
2301 using DPH into flat data parallelism. Flat data parallel programs
2302 should have better load balancing, enable SIMD parallelism and
2303 friendlier cache behaviour.</para>
2304 </listitem>
2305 </varlistentry>
2306
2307 <varlistentry>
2308 <term>
2309 <option>-favoid-vect</option>
2310 <indexterm><primary><option></option></primary></indexterm>
2311 </term>
2312 <listitem>
2313 <para>Part of <link linkend="dph">Data Parallel Haskell
2314 (DPH)</link>.</para>
2315
2316 <para><emphasis>Off by default.</emphasis> Enable the
2317 <emphasis>vectorisation</emphasis> avoidance optimisation. This
2318 optimisation only works when used in combination with the
2319 <option>-fvectorise</option> transformation.</para>
2320
2321 <para>While vectorisation of code using DPH is often a big win, it
2322 can also produce worse results for some kinds of code. This
2323 optimisation modifies the vectorisation transformation to try to
2324 determine if a function would be better of unvectorised and if
2325 so, do just that.</para>
2326 </listitem>
2327 </varlistentry>
2328
2329 <varlistentry>
2330 <term>
2331 <option>-fregs-graph</option>
2332 <indexterm><primary><option></option></primary></indexterm>
2333 </term>
2334 <listitem>
2335 <para><emphasis>Off by default, but enabled by -O2. Only applies in
2336 combination with the native code generator.</emphasis>
2337 Use the graph colouring register allocator for register allocation
2338 in the native code generator. By default, GHC uses a simpler,
2339 faster linear register allocator. The downside being that the
2340 linear register allocator usually generates worse code.
2341 </para>
2342 </listitem>
2343 </varlistentry>
2344
2345 <varlistentry>
2346 <term>
2347 <option>-fregs-iterative</option>
2348 <indexterm><primary><option></option></primary></indexterm>
2349 </term>
2350 <listitem>
2351 <para><emphasis>Off by default, only applies in combination with
2352 the native code generator.</emphasis>
2353 Use the iterative coalescing graph colouring register allocator for
2354 register allocation in the native code generator. This is the same
2355 register allocator as the <option>-freg-graph</option> one but also
2356 enables iterative coalescing during register allocation.
2357 </para>
2358 </listitem>
2359 </varlistentry>
2360
2361 <varlistentry>
2362 <term>
2363 <option>-fexcess-precision</option>
2364 <indexterm><primary><option>-fexcess-precision</option></primary></indexterm>
2365 </term>
2366 <listitem>
2367 <para>When this option is given, intermediate floating
2368 point values can have a <emphasis>greater</emphasis>
2369 precision/range than the final type. Generally this is a
2370 good thing, but some programs may rely on the exact
2371 precision/range of
2372 <literal>Float</literal>/<literal>Double</literal> values
2373 and should not use this option for their compilation.</para>
2374
2375 <para>
2376 Note that the 32-bit x86 native code generator only
2377 supports excess-precision mode, so neither
2378 <option>-fexcess-precision</option> nor
2379 <option>-fno-excess-precision</option> has any effect.
2380 This is a known bug, see <xref linkend="bugs-ghc" />.
2381 </para>
2382 </listitem>
2383 </varlistentry>
2384
2385 <varlistentry>
2386 <term>
2387 <option>-fignore-asserts</option>
2388 <indexterm><primary><option>-fignore-asserts</option></primary></indexterm>
2389 </term>
2390 <listitem>
2391 <para>Causes GHC to ignore uses of the function
2392 <literal>Exception.assert</literal> in source code (in
2393 other words, rewriting <literal>Exception.assert p
2394 e</literal> to <literal>e</literal> (see <xref
2395 linkend="assertions"/>). This flag is turned on by
2396 <option>-O</option>.
2397 </para>
2398 </listitem>
2399 </varlistentry>
2400
2401 <varlistentry>
2402 <term>
2403 <option>-fignore-interface-pragmas</option>
2404 <indexterm><primary><option>-fignore-interface-pragmas</option></primary></indexterm>
2405 </term>
2406 <listitem>
2407 <para>Tells GHC to ignore all inessential information when reading interface files.
2408 That is, even if <filename>M.hi</filename> contains unfolding or strictness information
2409 for a function, GHC will ignore that information.</para>
2410 </listitem>
2411 </varlistentry>
2412
2413 <varlistentry>
2414 <term>
2415 <option>-fomit-interface-pragmas</option>
2416 <indexterm><primary><option>-fomit-interface-pragmas</option></primary></indexterm>
2417 </term>
2418 <listitem>
2419 <para>Tells GHC to omit all inessential information from the
2420 interface file generated for the module being compiled (say M).
2421 This means that a module importing M will see only the
2422 <emphasis>types</emphasis> of the functions that M exports, but
2423 not their unfoldings, strictness info, etc. Hence, for example,
2424 no function exported by M will be inlined into an importing module.
2425 The benefit is that modules that import M will need to be
2426 recompiled less often (only when M's exports change their type, not
2427 when they change their implementation).</para>
2428 </listitem>
2429 </varlistentry>
2430
2431 <varlistentry>
2432 <term>
2433 <option>-fomit-yields</option>
2434 <indexterm><primary><option>-fomit-yields</option></primary></indexterm>
2435 </term>
2436 <listitem>
2437 <para><emphasis>On by default.</emphasis> Tells GHC to omit
2438 heap checks when no allocation is being performed. While this improves
2439 binary sizes by about 5%, it also means that threads run in
2440 tight non-allocating loops will not get preempted in a timely
2441 fashion. If it is important to always be able to interrupt such
2442 threads, you should turn this optimization off. Consider also
2443 recompiling all libraries with this optimization turned off, if you
2444 need to guarantee interruptibility.
2445 </para>
2446 </listitem>
2447 </varlistentry>
2448
2449 </variablelist>
2450
2451 </sect2>
2452
2453 </sect1>
2454
2455 &code-gens;
2456
2457 &phases;
2458
2459 &shared_libs;
2460
2461 <sect1 id="using-concurrent">
2462 <title>Using Concurrent Haskell</title>
2463 <indexterm><primary>Concurrent Haskell</primary><secondary>using</secondary></indexterm>
2464
2465 <para>GHC supports Concurrent Haskell by default, without requiring a
2466 special option or libraries compiled in a certain way. To get access to
2467 the support libraries for Concurrent Haskell, just import
2468 <ulink
2469 url="&libraryBaseLocation;/Control-Concurrent.html"><literal>Control.Concurrent</literal></ulink>. More information on Concurrent Haskell is provided in the documentation for that module.</para>
2470
2471 <para>
2472 Optionally, the program may be linked with
2473 the <option>-threaded</option> option (see
2474 <xref linkend="options-linker" />. This provides two benefits:
2475
2476 <itemizedlist>
2477 <listitem>
2478 <para>It enables the <option>-N</option><indexterm><primary><option>-N<replaceable>x</replaceable></option></primary><secondary>RTS option</secondary></indexterm> RTS option to be
2479 used, which allows threads to run in
2480 parallel<indexterm><primary>parallelism</primary></indexterm>
2481 on a
2482 multiprocessor<indexterm><primary>multiprocessor</primary></indexterm><indexterm><primary>SMP</primary></indexterm>
2483 or
2484 multicore<indexterm><primary>multicore</primary></indexterm>
2485 machine. See <xref linkend="using-smp" />.</para>
2486 </listitem>
2487 <listitem>
2488 <para>If a thread makes a foreign call (and the call is
2489 not marked <literal>unsafe</literal>), then other
2490 Haskell threads in the program will continue to run
2491 while the foreign call is in progress.
2492 Additionally, <literal>foreign export</literal>ed
2493 Haskell functions may be called from multiple OS
2494 threads simultaneously. See
2495 <xref linkend="ffi-threads" />.</para>
2496 </listitem>
2497 </itemizedlist>
2498 </para>
2499
2500 <para>The following RTS option(s) affect the behaviour of Concurrent
2501 Haskell programs:<indexterm><primary>RTS options, concurrent</primary></indexterm></para>
2502
2503 <variablelist>
2504 <varlistentry>
2505 <term><option>-C<replaceable>s</replaceable></option></term>
2506 <listitem>
2507 <para><indexterm><primary><option>-C<replaceable>s</replaceable></option></primary><secondary>RTS option</secondary></indexterm>
2508 Sets the context switch interval to <replaceable>s</replaceable>
2509 seconds. A context switch will occur at the next heap block
2510 allocation after the timer expires (a heap block allocation occurs
2511 every 4k of allocation). With <option>-C0</option> or
2512 <option>-C</option>, context switches will occur as often as
2513 possible (at every heap block allocation). By default, context
2514 switches occur every 20ms.</para>
2515 </listitem>
2516 </varlistentry>
2517 </variablelist>
2518 </sect1>
2519
2520 <sect1 id="using-smp">
2521 <title>Using SMP parallelism</title>
2522 <indexterm><primary>parallelism</primary>
2523 </indexterm>
2524 <indexterm><primary>SMP</primary>
2525 </indexterm>
2526
2527 <para>GHC supports running Haskell programs in parallel on an SMP
2528 (symmetric multiprocessor).</para>
2529
2530 <para>There's a fine distinction between
2531 <emphasis>concurrency</emphasis> and <emphasis>parallelism</emphasis>:
2532 parallelism is all about making your program run
2533 <emphasis>faster</emphasis> by making use of multiple processors
2534 simultaneously. Concurrency, on the other hand, is a means of
2535 abstraction: it is a convenient way to structure a program that must
2536 respond to multiple asynchronous events.</para>
2537
2538 <para>However, the two terms are certainly related. By making use of
2539 multiple CPUs it is possible to run concurrent threads in parallel,
2540 and this is exactly what GHC's SMP parallelism support does. But it
2541 is also possible to obtain performance improvements with parallelism
2542 on programs that do not use concurrency. This section describes how to
2543 use GHC to compile and run parallel programs, in <xref
2544 linkend="lang-parallel" /> we describe the language features that affect
2545 parallelism.</para>
2546
2547 <sect2 id="parallel-compile-options">
2548 <title>Compile-time options for SMP parallelism</title>
2549
2550 <para>In order to make use of multiple CPUs, your program must be
2551 linked with the <option>-threaded</option> option (see <xref
2552 linkend="options-linker" />). Additionally, the following
2553 compiler options affect parallelism:</para>
2554
2555 <variablelist>
2556 <varlistentry>
2557 <term><option>-feager-blackholing</option></term>
2558 <indexterm><primary><option>-feager-blackholing</option></primary></indexterm>
2559 <listitem>
2560 <para>
2561 Blackholing is the act of marking a thunk (lazy
2562 computuation) as being under evaluation. It is useful for
2563 three reasons: firstly it lets us detect certain kinds of
2564 infinite loop (the <literal>NonTermination</literal>
2565 exception), secondly it avoids certain kinds of space
2566 leak, and thirdly it avoids repeating a computation in a
2567 parallel program, because we can tell when a computation
2568 is already in progress.</para>
2569
2570 <para>
2571 The option <option>-feager-blackholing</option> causes
2572 each thunk to be blackholed as soon as evaluation begins.
2573 The default is "lazy blackholing", whereby thunks are only
2574 marked as being under evaluation when a thread is paused
2575 for some reason. Lazy blackholing is typically more
2576 efficient (by 1-2&percnt; or so), because most thunks don't
2577 need to be blackholed. However, eager blackholing can
2578 avoid more repeated computation in a parallel program, and
2579 this often turns out to be important for parallelism.
2580 </para>
2581
2582 <para>
2583 We recommend compiling any code that is intended to be run
2584 in parallel with the <option>-feager-blackholing</option>
2585 flag.
2586 </para>
2587 </listitem>
2588 </varlistentry>
2589 </variablelist>
2590 </sect2>
2591
2592 <sect2 id="parallel-options">
2593 <title>RTS options for SMP parallelism</title>
2594
2595 <para>There are two ways to run a program on multiple
2596 processors:
2597 call <literal>Control.Concurrent.setNumCapabilities</literal> from your
2598 program, or use the RTS <option>-N</option> option.</para>
2599
2600 <variablelist>
2601 <varlistentry>
2602 <term><option>-N<optional><replaceable>x</replaceable></optional></option></term>
2603 <listitem>
2604 <para><indexterm><primary><option>-N<replaceable>x</replaceable></option></primary><secondary>RTS option</secondary></indexterm>
2605 Use <replaceable>x</replaceable> simultaneous threads when
2606 running the program. Normally <replaceable>x</replaceable>
2607 should be chosen to match the number of CPU cores on the
2608 machine<footnote><para>Whether hyperthreading cores should be counted or not is an
2609 open question; please feel free to experiment and let us know what
2610 results you find.</para></footnote>. For example,
2611 on a dual-core machine we would probably use
2612 <literal>+RTS -N2 -RTS</literal>.</para>
2613
2614 <para>Omitting <replaceable>x</replaceable>,
2615 i.e. <literal>+RTS -N -RTS</literal>, lets the runtime
2616 choose the value of <replaceable>x</replaceable> itself
2617 based on how many processors are in your machine.</para>
2618
2619 <para>Be careful when using all the processors in your
2620 machine: if some of your processors are in use by other
2621 programs, this can actually harm performance rather than
2622 improve it.</para>
2623
2624 <para>Setting <option>-N</option> also has the effect of
2625 enabling the parallel garbage collector (see
2626 <xref linkend="rts-options-gc" />).</para>
2627
2628 <para>The current value of the <option>-N</option> option
2629 is available to the Haskell program
2630 via <literal>Control.Concurrent.getNumCapabilities</literal>, and
2631 it may be changed while the program is running by
2632 calling <literal>Control.Concurrent.setNumCapabilities</literal>.</para>
2633 </listitem>
2634 </varlistentry>
2635 </variablelist>
2636
2637 <para>The following options affect the way the runtime schedules
2638 threads on CPUs:</para>
2639
2640 <variablelist>
2641 <varlistentry>
2642 <term><option>-qa</option></term>
2643 <indexterm><primary><option>-qa</option></primary><secondary>RTS
2644 option</secondary></indexterm>
2645 <listitem>
2646 <para>Use the OS's affinity facilities to try to pin OS
2647 threads to CPU cores. This is an experimental feature,
2648 and may or may not be useful. Please let us know
2649 whether it helps for you!</para>
2650 </listitem>
2651 </varlistentry>
2652 <varlistentry>
2653 <term><option>-qm</option></term>
2654 <indexterm><primary><option>-qm</option></primary><secondary>RTS
2655 option</secondary></indexterm>
2656 <listitem>
2657 <para>Disable automatic migration for load balancing.
2658 Normally the runtime will automatically try to schedule
2659 threads across the available CPUs to make use of idle
2660 CPUs; this option disables that behaviour. Note that
2661 migration only applies to threads; sparks created
2662 by <literal>par</literal> are load-balanced separately
2663 by work-stealing.</para>
2664
2665 <para>
2666 This option is probably only of use for concurrent
2667 programs that explicitly schedule threads onto CPUs
2668 with <literal>Control.Concurrent.forkOn</literal>.
2669 </para>
2670 </listitem>
2671 </varlistentry>
2672 </variablelist>
2673 </sect2>
2674
2675 <sect2>
2676 <title>Hints for using SMP parallelism</title>
2677
2678 <para>Add the <literal>-s</literal> RTS option when
2679 running the program to see timing stats, which will help to tell you
2680 whether your program got faster by using more CPUs or not. If the user
2681 time is greater than
2682 the elapsed time, then the program used more than one CPU. You should
2683 also run the program without <literal>-N</literal> for
2684 comparison.</para>
2685
2686 <para>The output of <literal>+RTS -s</literal> tells you how
2687 many &ldquo;sparks&rdquo; were created and executed during the
2688 run of the program (see <xref linkend="rts-options-gc" />), which
2689 will give you an idea how well your <literal>par</literal>
2690 annotations are working.</para>
2691
2692 <para>GHC's parallelism support has improved in 6.12.1 as a
2693 result of much experimentation and tuning in the runtime
2694 system. We'd still be interested to hear how well it works
2695 for you, and we're also interested in collecting parallel
2696 programs to add to our benchmarking suite.</para>
2697 </sect2>
2698 </sect1>
2699
2700 <sect1 id="options-platform">
2701 <title>Platform-specific Flags</title>
2702
2703 <indexterm><primary>-m* options</primary></indexterm>
2704 <indexterm><primary>platform-specific options</primary></indexterm>
2705 <indexterm><primary>machine-specific options</primary></indexterm>
2706
2707 <para>Some flags only make sense for particular target
2708 platforms.</para>
2709
2710 <variablelist>
2711
2712 <varlistentry>
2713 <term><option>-msse2</option>:</term>
2714 <listitem>
2715 <para>
2716 (x86 only, added in GHC 7.0.1) Use the SSE2 registers and
2717 instruction set to implement floating point operations when using
2718 the <link linkend="native-code-gen">native code generator</link>.
2719 This gives a substantial performance improvement for floating
2720 point, but the resulting compiled code
2721 will only run on processors that support SSE2 (Intel Pentium 4 and
2722 later, or AMD Athlon 64 and later). The
2723 <link linkend="llvm-code-gen">LLVM backend</link> will also use SSE2
2724 if your processor supports it but detects this automatically so no
2725 flag is required.
2726 </para>
2727 <para>
2728 SSE2 is unconditionally used on x86-64 platforms.
2729 </para>
2730 </listitem>
2731 </varlistentry>
2732
2733 <varlistentry>
2734 <term><option>-msse4.2</option>:</term>
2735 <listitem>
2736 <para>
2737 (x86 only, added in GHC 7.4.1) Use the SSE4.2 instruction set to
2738 implement some floating point and bit operations when using the
2739 <link linkend="native-code-gen">native code generator</link>. The
2740 resulting compiled code will only run on processors that
2741 support SSE4.2 (Intel Core i7 and later). The
2742 <link linkend="llvm-code-gen">LLVM backend</link> will also use
2743 SSE4.2 if your processor supports it but detects this automatically
2744 so no flag is required.
2745 </para>
2746 </listitem>
2747 </varlistentry>
2748
2749 </variablelist>
2750
2751 </sect1>
2752
2753 &runtime;
2754
2755 <sect1 id="ext-core">
2756 <title>Generating and compiling External Core Files</title>
2757
2758 <indexterm><primary>intermediate code generation</primary></indexterm>
2759
2760 <para>GHC can dump its optimized intermediate code (said to be in &ldquo;Core&rdquo; format)
2761 to a file as a side-effect of compilation. Non-GHC back-end tools can read and process Core files; these files have the suffix
2762 <filename>.hcr</filename>. The Core format is described in <ulink url="../../core.pdf">
2763 <citetitle>An External Representation for the GHC Core Language</citetitle></ulink>,
2764 and sample tools
2765 for manipulating Core files (in Haskell) are available in the
2766 <ulink url="http://hackage.haskell.org/package/extcore">extcore package on Hackage</ulink>. Note that the format of <literal>.hcr</literal>
2767 files is <emphasis>different</emphasis> from the Core output format that GHC generates
2768 for debugging purposes (<xref linkend="options-debugging"/>), though the two formats appear somewhat similar.</para>
2769
2770 <para>The Core format natively supports notes which you can add to
2771 your source code using the <literal>CORE</literal> pragma (see <xref
2772 linkend="pragmas"/>).</para>
2773
2774 <variablelist>
2775
2776 <varlistentry>
2777 <term>
2778 <option>-fext-core</option>
2779 <indexterm><primary><option>-fext-core</option></primary></indexterm>
2780 </term>
2781 <listitem>
2782 <para>Generate <literal>.hcr</literal> files.</para>
2783 </listitem>
2784 </varlistentry>
2785
2786 </variablelist>
2787
2788 <para>Currently (as of version 6.8.2), GHC does not have the ability to read in External Core files as source. If you would like GHC to have this ability, please <ulink url="http://hackage.haskell.org/trac/ghc/wiki/MailingListsAndIRC">make your wishes known to the GHC Team</ulink>.</para>
2789
2790 </sect1>
2791
2792 &debug;
2793 &flags;
2794
2795 </chapter>
2796
2797 <!-- Emacs stuff:
2798 ;;; Local Variables: ***
2799 ;;; sgml-parent-document: ("users_guide.xml" "book" "chapter") ***
2800 ;;; End: ***
2801 -->