1 <?xml version="1.0" encoding="iso-8859-1"?>
2 <chapter id="using-ghc">
3 <title>Using GHC</title>
5 <indexterm><primary>GHC, using</primary></indexterm>
6 <indexterm><primary>using GHC</primary></indexterm>
9 <title>Options overview</title>
11 <para>GHC's behaviour is controlled by
12 <firstterm>options</firstterm>, which for historical reasons are
13 also sometimes referred to as command-line flags or arguments.
14 Options can be specified in three ways:</para>
17 <title>command-line arguments</title>
19 <indexterm><primary>structure, command-line</primary></indexterm>
20 <indexterm><primary>command-line</primary><secondary>arguments</secondary></indexterm>
21 <indexterm><primary>arguments</primary><secondary>command-line</secondary></indexterm>
23 <para>An invocation of GHC takes the following form:</para>
29 <para>command-line arguments are either options or file names.</para>
31 <para>command-line options begin with <literal>-</literal>.
32 They may <emphasis>not</emphasis> be grouped:
33 <option>-vO</option> is different from <option>-v -O</option>.
34 Options need not precede filenames: e.g., <literal>ghc *.o -o
35 foo</literal>. All options are processed and then applied to
36 all files; you cannot, for example, invoke <literal>ghc -c -O1
37 Foo.hs -O2 Bar.hs</literal> to apply different optimisation
38 levels to the files <filename>Foo.hs</filename> and
39 <filename>Bar.hs</filename>.</para>
42 <sect2 id="source-file-options">
43 <title>command line options in source files</title>
45 <indexterm><primary>source-file options</primary></indexterm>
47 <para>Sometimes it is useful to make the connection between a
48 source file and the command-line options it requires quite
49 tight. For instance, if a Haskell source file uses GHC
50 extensions, it will always need to be compiled with the
51 <option>-fglasgow-exts</option> option. Rather than maintaining
52 the list of per-file options in a <filename>Makefile</filename>,
53 it is possible to do this directly in the source file using the
54 <literal>OPTIONS_GHC</literal> pragma <indexterm><primary>OPTIONS_GHC
55 pragma</primary></indexterm>:</para>
58 {-# OPTIONS_GHC -fglasgow-exts #-}
63 <para><literal>OPTIONS_GHC</literal> pragmas are only looked for at
64 the top of your source files, upto the first
65 (non-literate,non-empty) line not containing
66 <literal>OPTIONS_GHC</literal>. Multiple <literal>OPTIONS_GHC</literal>
67 pragmas are recognised. Do not put comments before, or on the same line
68 as, the <literal>OPTIONS_GHC</literal> pragma.</para>
70 <para>Note that your command shell does not
71 get to the source file options, they are just included literally
72 in the array of command-line arguments the compiler
73 maintains internally, so you'll be desperately disappointed if
74 you try to glob etc. inside <literal>OPTIONS_GHC</literal>.</para>
76 <para>NOTE: the contents of OPTIONS_GHC are prepended to the
77 command-line options, so you <emphasis>do</emphasis> have the
78 ability to override OPTIONS_GHC settings via the command
81 <para>It is not recommended to move all the contents of your
82 Makefiles into your source files, but in some circumstances, the
83 <literal>OPTIONS_GHC</literal> pragma is the Right Thing. (If you
84 use <option>-keep-hc-file-too</option> and have OPTION flags in
85 your module, the OPTIONS_GHC will get put into the generated .hc
90 <title>Setting options in GHCi</title>
92 <para>Options may also be modified from within GHCi, using the
93 <literal>:set</literal> command. See <xref linkend="ghci-set"/>
94 for more details.</para>
98 <sect1 id="static-dynamic-flags">
99 <title>Static vs. Dynamic options</title>
100 <indexterm><primary>static</primary><secondary>options</secondary>
102 <indexterm><primary>dynamic</primary><secondary>options</secondary>
105 <para>Each of GHC's command line options is classified as either
106 <firstterm>static</firstterm> or <firstterm>dynamic</firstterm>.
107 A static flag may only be specified on the command line, whereas a
108 dynamic flag may also be given in an <literal>OPTIONS_GHC</literal>
109 pragma in a source file or set from the GHCi command-line with
110 <literal>:set</literal>.</para>
112 <para>As a rule of thumb, options which relate to filenames are
113 static, and the rest are dynamic. The flag reference tables (<xref
114 linkend="flag-reference"/>) lists the status of each flag.</para>
117 <sect1 id="file-suffixes">
118 <title>Meaningful file suffixes</title>
120 <indexterm><primary>suffixes, file</primary></indexterm>
121 <indexterm><primary>file suffixes for GHC</primary></indexterm>
123 <para>File names with “meaningful” suffixes (e.g.,
124 <filename>.lhs</filename> or <filename>.o</filename>) cause the
125 “right thing” to happen to those files.</para>
131 <filename>.lhs</filename>
132 <indexterm><primary><literal>lhs</literal> suffix</primary></indexterm>
135 <para>A “literate Haskell” module.</para>
140 <term><filename>.hs</filename></term>
142 <para>A not-so-literate Haskell module.</para>
147 <term><filename>.hi</filename></term>
149 <para>A Haskell interface file, probably
150 compiler-generated.</para>
155 <term><filename>.hc</filename></term>
157 <para>Intermediate C file produced by the Haskell
163 <term><filename>.c</filename></term>
165 <para>A C file not produced by the Haskell
171 <term><filename>.s</filename></term>
173 <para>An assembly-language source file, usually produced by
179 <term><filename>.o</filename></term>
181 <para>An object file, produced by an assembler.</para>
186 <para>Files with other suffixes (or without suffixes) are passed
187 straight to the linker.</para>
192 <title>Modes of operation</title>
194 <para>GHC's behaviour is firstly controlled by a mode flag. Only
195 one of these flags may be given, but it does not necessarily need
196 to be the first option on the command-line. The available modes
202 <cmdsynopsis><command>ghc</command>
203 <arg choice='plain'>––interactive</arg>
205 <indexterm><primary>interactive mode</primary></indexterm>
206 <indexterm><primary>ghci</primary></indexterm>
209 <para>Interactive mode, which is also available as
210 <command>ghci</command>. Interactive mode is described in
211 more detail in <xref linkend="ghci"/>.</para>
217 <cmdsynopsis><command>ghc</command>
218 <arg choice='plain'>––make</arg>
220 <indexterm><primary>make mode</primary></indexterm>
221 <indexterm><primary><option>––make</option></primary></indexterm>
224 <para>In this mode, GHC will build a multi-module Haskell
225 program automatically, figuring out dependencies for itself.
226 If you have a straightforward Haskell program, this is
227 likely to be much easier, and faster, than using
228 <command>make</command>. Make mode is described in <xref
229 linkend="make-mode"/>.</para>
235 <cmdsynopsis><command>ghc</command>
236 <arg choice='plain'>–e</arg> <arg choice='plain'><replaceable>expr</replaceable></arg>
238 <indexterm><primary>eval mode</primary></indexterm>
241 <para>Expression-evaluation mode. This is very similar to
242 interactive mode, except that there is a single expression
243 to evaluate (<replaceable>expr</replaceable>) which is given
244 on the command line. See <xref linkend="eval-mode"/> for
252 <command>ghc</command>
260 <indexterm><primary><option>-E</option></primary></indexterm>
261 <indexterm><primary><option>-C</option></primary></indexterm>
262 <indexterm><primary><option>-S</option></primary></indexterm>
263 <indexterm><primary><option>-c</option></primary></indexterm>
266 <para>This is the traditional batch-compiler mode, in which
267 GHC can compile source files one at a time, or link objects
268 together into an executable. This mode also applies if
269 there is no other mode flag specified on the command line,
270 in which case it means that the specified files should be
271 compiled and then linked to form a program. See <xref
272 linkend="options-order"/>.</para>
279 <command>ghc</command>
280 <arg choice='plain'>–M</arg>
282 <indexterm><primary>dependency-generation mode</primary></indexterm>
285 <para>Dependency-generation mode. In this mode, GHC can be
286 used to generate dependency information suitable for use in
287 a <literal>Makefile</literal>. See <xref
288 linkend="sec-makefile-dependencies"/>.</para>
295 <command>ghc</command>
296 <arg choice='plain'>––mk-dll</arg>
298 <indexterm><primary>dependency-generation mode</primary></indexterm>
301 <para>DLL-creation mode (Windows only). See <xref
302 linkend="win32-dlls-create"/>.</para>
307 <sect2 id="make-mode">
308 <title>Using <command>ghc</command> <option>––make</option></title>
309 <indexterm><primary><option>––make</option></primary></indexterm>
310 <indexterm><primary>separate compilation</primary></indexterm>
312 <para>When given the <option>––make</option> option,
313 GHC will build a multi-module Haskell program by following
314 dependencies from a single root module (usually
315 <literal>Main</literal>). For example, if your
316 <literal>Main</literal> module is in a file called
317 <filename>Main.hs</filename>, you could compile and link the
318 program like this:</para>
321 ghc ––make Main.hs
324 <para>The command line may contain any number of source file
325 names or module names; GHC will figure out all the modules in
326 the program by following the imports from these initial modules.
327 It will then attempt to compile each module which is out of
328 date, and finally if there is a <literal>Main</literal> module,
329 the program will also be linked into an executable.</para>
331 <para>The main advantages to using <literal>ghc
332 ––make</literal> over traditional
333 <literal>Makefile</literal>s are:</para>
337 <para>GHC doesn't have to be restarted for each compilation,
338 which means it can cache information between compilations.
339 Compiling a multi-module program with <literal>ghc
340 ––make</literal> can be up to twice as fast as
341 running <literal>ghc</literal> individually on each source
345 <para>You don't have to write a<literal>Makefile</literal>.</para>
346 <indexterm><primary><literal>Makefile</literal>s</primary><secondary>avoiding</secondary></indexterm>
349 <para>GHC re-calculates the dependencies each time it is
350 invoked, so the dependencies never get out of sync with the
355 <para>Any of the command-line options described in the rest of
356 this chapter can be used with
357 <option>––make</option>, but note that any options
358 you give on the command line will apply to all the source files
359 compiled, so if you want any options to apply to a single source
360 file only, you'll need to use an <literal>OPTIONS_GHC</literal>
361 pragma (see <xref linkend="source-file-options"/>).</para>
363 <para>If the program needs to be linked with additional objects
364 (say, some auxiliary C code), then the object files can be
365 given on the command line and GHC will include them when linking
366 the executable.</para>
368 <para>Note that GHC can only follow dependencies if it has the
369 source file available, so if your program includes a module for
370 which there is no source file, even if you have an object and an
371 interface file for the module, then GHC will complain. The
372 exception to this rule is for package modules, which may or may
373 not have source files.</para>
375 <para>The source files for the program don't all need to be in
376 the same directory; the <option>-i</option> option can be used
377 to add directories to the search path (see <xref
378 linkend="search-path"/>).</para>
381 <sect2 id="eval-mode">
382 <title>Expression evaluation mode</title>
384 <para>This mode is very similar to interactive mode, except that
385 there is a single expression to evaluate which is specified on
386 the command line as an argument to the <option>-e</option>
390 ghc -e <replaceable>expr</replaceable>
393 <para>Haskell source files may be named on the command line, and
394 they will be loaded exactly as in interactive mode. The
395 expression is evaluated in the context of the loaded
398 <para>For example, to load and run a Haskell program containing
399 a module <literal>Main</literal>, we might say</para>
402 ghc -e Main.main Main.hs
405 <para>or we can just use this mode to evaluate expressions in
406 the context of the <literal>Prelude</literal>:</para>
409 $ ghc -e "interact (unlines.map reverse.lines)"
415 <sect2 id="options-order">
416 <title>Batch compiler mode</title>
418 <para>In <emphasis>batch mode</emphasis>, GHC will compile one or more source files
419 given on the command line.</para>
421 <para>The first phase to run is determined by each input-file
422 suffix, and the last phase is determined by a flag. If no
423 relevant flag is present, then go all the way through linking.
424 This table summarises:</para>
428 <colspec align="left"/>
429 <colspec align="left"/>
430 <colspec align="left"/>
431 <colspec align="left"/>
435 <entry>Phase of the compilation system</entry>
436 <entry>Suffix saying “start here”</entry>
437 <entry>Flag saying “stop after”</entry>
438 <entry>(suffix of) output file</entry>
443 <entry>literate pre-processor</entry>
444 <entry><literal>.lhs</literal></entry>
446 <entry><literal>.hs</literal></entry>
450 <entry>C pre-processor (opt.) </entry>
451 <entry><literal>.hs</literal> (with
452 <option>-cpp</option>)</entry>
453 <entry><option>-E</option></entry>
454 <entry><literal>.hspp</literal></entry>
458 <entry>Haskell compiler</entry>
459 <entry><literal>.hs</literal></entry>
460 <entry><option>-C</option>, <option>-S</option></entry>
461 <entry><literal>.hc</literal>, <literal>.s</literal></entry>
465 <entry>C compiler (opt.)</entry>
466 <entry><literal>.hc</literal> or <literal>.c</literal></entry>
467 <entry><option>-S</option></entry>
468 <entry><literal>.s</literal></entry>
472 <entry>assembler</entry>
473 <entry><literal>.s</literal></entry>
474 <entry><option>-c</option></entry>
475 <entry><literal>.o</literal></entry>
479 <entry>linker</entry>
480 <entry><replaceable>other</replaceable></entry>
482 <entry><filename>a.out</filename></entry>
488 <indexterm><primary><option>-C</option></primary></indexterm>
489 <indexterm><primary><option>-E</option></primary></indexterm>
490 <indexterm><primary><option>-S</option></primary></indexterm>
491 <indexterm><primary><option>-c</option></primary></indexterm>
493 <para>Thus, a common invocation would be: </para>
496 ghc -c Foo.hs</screen>
498 <para>to compile the Haskell source file
499 <filename>Foo.hs</filename> to an object file
500 <filename>Foo.o</filename>.</para>
502 <para>Note: What the Haskell compiler proper produces depends on
503 whether a native-code generator<indexterm><primary>native-code
504 generator</primary></indexterm> is used (producing assembly
505 language) or not (producing C). See <xref
506 linkend="options-codegen"/> for more details.</para>
508 <para>Note: C pre-processing is optional, the
509 <option>-cpp</option><indexterm><primary><option>-cpp</option></primary></indexterm>
510 flag turns it on. See <xref linkend="c-pre-processor"/> for more
513 <para>Note: The option <option>-E</option><indexterm><primary>-E
514 option</primary></indexterm> runs just the pre-processing passes
515 of the compiler, dumping the result in a file. Note that this
516 differs from the previous behaviour of dumping the file to
517 standard output.</para>
519 <sect3 id="overriding-suffixes">
520 <title>Overriding the default behaviour for a file</title>
522 <para>As described above, the way in which a file is processed by GHC
523 depends on its suffix. This behaviour can be overriden using the
524 <option>-x</option> option:</para>
528 <term><option>-x</option> <replaceable>suffix</replaceable></term>
529 <indexterm><primary><option>-x</option></primary>
532 <para>Causes all files following this option on the command
533 line to be processed as if they had the suffix
534 <replaceable>suffix</replaceable>. For example, to compile a
535 Haskell module in the file <literal>M.my-hs</literal>,
536 use <literal>ghc -c -x hs M.my-hs</literal>.</para>
545 <sect1 id="options-help">
546 <title>Help and verbosity options</title>
548 <indexterm><primary>help options</primary></indexterm>
549 <indexterm><primary>verbosity options</primary></indexterm>
554 <option>––help</option>
555 <indexterm><primary><option>––help</option></primary></indexterm>
559 <indexterm><primary><option>-?</option></primary></indexterm>
562 <para>Cause GHC to spew a long usage message to standard
563 output and then exit.</para>
570 <indexterm><primary><option>-v</option></primary></indexterm>
573 <para>The <option>-v</option> option makes GHC
574 <emphasis>verbose</emphasis>: it reports its version number
575 and shows (on stderr) exactly how it invokes each phase of
576 the compilation system. Moreover, it passes the
577 <option>-v</option> flag to most phases; each reports its
578 version number (and possibly some other information).</para>
580 <para>Please, oh please, use the <option>-v</option> option
581 when reporting bugs! Knowing that you ran the right bits in
582 the right order is always the first thing we want to
589 <option>-v</option><replaceable>n</replaceable>
590 <indexterm><primary><option>-v</option></primary></indexterm>
593 <para>To provide more control over the compiler's verbosity,
594 the <option>-v</option> flag takes an optional numeric
595 argument. Specifying <option>-v</option> on its own is
596 equivalent to <option>-v3</option>, and the other levels
597 have the following meanings:</para>
601 <term><option>-v0</option></term>
603 <para>Disable all non-essential messages (this is the
609 <term><option>-v1</option></term>
611 <para>Minimal verbosity: print one line per
612 compilation (this is the default when
613 <option>––make</option> or
614 <option>––interactive</option> is on).</para>
619 <term><option>-v2</option></term>
621 <para>Print the name of each compilation phase as it
622 is executed. (equivalent to
623 <option>-dshow-passes</option>).</para>
628 <term><option>-v3</option></term>
630 <para>The same as <option>-v2</option>, except that in
631 addition the full command line (if appropriate) for
632 each compilation phase is also printed.</para>
637 <term><option>-v4</option></term>
639 <para>The same as <option>-v3</option> except that the
640 intermediate program representation after each
641 compilation phase is also printed (excluding
642 preprocessed and C/assembly files).</para>
652 <indexterm><primary><option>-V</option></primary></indexterm>
655 <option>––version</option>
656 <indexterm><primary><option>––version</option></primary></indexterm>
659 <para>Print a one-line string including GHC's version number.</para>
665 <option>––numeric-version</option>
666 <indexterm><primary><option>––numeric-version</option></primary></indexterm>
669 <para>Print GHC's numeric version number only.</para>
675 <option>––print-libdir</option>
676 <indexterm><primary><option>––print-libdir</option></primary></indexterm>
679 <para>Print the path to GHC's library directory. This is
680 the top of the directory tree containing GHC's libraries,
681 interfaces, and include files (usually something like
682 <literal>/usr/local/lib/ghc-5.04</literal> on Unix). This
684 <literal>$libdir</literal><indexterm><primary><literal>libdir</literal></primary>
685 </indexterm>in the package configuration file (see <xref
686 linkend="packages"/>).</para>
691 <term><option>-ferror-spans</option>
692 <indexterm><primary><option>-ferror-spans</option></primary>
696 <para>Causes GHC to emit the full source span of the
697 syntactic entity relating to an error message. Normally, GHC
698 emits the source location of the start of the syntactic
701 <para>For example:</para>
703 <screen>test.hs:3:6: parse error on input `where'</screen>
705 <para>becomes:</para>
707 <screen>test296.hs:3:6-10: parse error on input `where'</screen>
709 <para>And multi-line spans are possible too:</para>
711 <screen>test.hs:(5,4)-(6,7):
712 Conflicting definitions for `a'
713 Bound at: test.hs:5:4
715 In the binding group for: a, b, a</screen>
717 <para>Note that line numbers start counting at one, but
718 column numbers start at zero. This choice was made to
719 follow existing convention (i.e. this is how Emacs does
728 <sect1 id="options-sanity">
729 <title>Warnings and sanity-checking</title>
731 <indexterm><primary>sanity-checking options</primary></indexterm>
732 <indexterm><primary>warnings</primary></indexterm>
735 <para>GHC has a number of options that select which types of
736 non-fatal error messages, otherwise known as warnings, can be
737 generated during compilation. By default, you get a standard set
738 of warnings which are generally likely to indicate bugs in your
740 <option>-fwarn-overlapping-patterns</option>,
741 <option>-fwarn-deprecations</option>,
742 <option>-fwarn-duplicate-exports</option>,
743 <option>-fwarn-missing-fields</option>, and
744 <option>-fwarn-missing-methods</option>. The following flags are
745 simple ways to select standard “packages” of warnings:
751 <term><option>-W</option>:</term>
753 <indexterm><primary>-W option</primary></indexterm>
754 <para>Provides the standard warnings plus
755 <option>-fwarn-incomplete-patterns</option>,
756 <option>-fwarn-unused-matches</option>,
757 <option>-fwarn-unused-imports</option>,
758 <option>-fwarn-misc</option>, and
759 <option>-fwarn-unused-binds</option>.</para>
764 <term><option>-w</option>:</term>
766 <indexterm><primary><option>-w</option></primary></indexterm>
767 <para>Turns off all warnings, including the standard ones.</para>
772 <term><option>-Wall</option>:</term>
774 <indexterm><primary><option>-Wall</option></primary></indexterm>
775 <para>Turns on all warning options.</para>
780 <term><option>-Werror</option>:</term>
782 <indexterm><primary><option>-Werror</option></primary></indexterm>
783 <para>Makes any warning into a fatal error. Useful so that you don't
784 miss warnings when doing batch compilation. </para>
790 <para>The full set of warning options is described below. To turn
791 off any warning, simply give the corresponding
792 <option>-fno-warn-...</option> option on the command line.</para>
797 <term><option>-fwarn-deprecations</option>:</term>
799 <indexterm><primary><option>-fwarn-deprecations</option></primary>
801 <indexterm><primary>deprecations</primary></indexterm>
802 <para>Causes a warning to be emitted when a deprecated
803 function or type is used. Entities can be marked as
804 deprecated using a pragma, see <xref
805 linkend="deprecated-pragma"/>.</para>
810 <term><option>-fwarn-duplicate-exports</option>:</term>
812 <indexterm><primary><option>-fwarn-duplicate-exports</option></primary></indexterm>
813 <indexterm><primary>duplicate exports, warning</primary></indexterm>
814 <indexterm><primary>export lists, duplicates</primary></indexterm>
816 <para>Have the compiler warn about duplicate entries in
817 export lists. This is useful information if you maintain
818 large export lists, and want to avoid the continued export
819 of a definition after you've deleted (one) mention of it in
820 the export list.</para>
822 <para>This option is on by default.</para>
827 <term><option>-fwarn-hi-shadowing</option>:</term>
829 <indexterm><primary><option>-fwarn-hi-shadowing</option></primary></indexterm>
830 <indexterm><primary>shadowing</primary>
831 <secondary>interface files</secondary></indexterm>
833 <para>Causes the compiler to emit a warning when a module or
834 interface file in the current directory is shadowing one
835 with the same module name in a library or other
841 <term><option>-fwarn-incomplete-patterns</option>:</term>
843 <indexterm><primary><option>-fwarn-incomplete-patterns</option></primary></indexterm>
844 <indexterm><primary>incomplete patterns, warning</primary></indexterm>
845 <indexterm><primary>patterns, incomplete</primary></indexterm>
847 <para>Similarly for incomplete patterns, the function
848 <function>g</function> below will fail when applied to
849 non-empty lists, so the compiler will emit a warning about
850 this when <option>-fwarn-incomplete-patterns</option> is
857 <para>This option isn't enabled be default because it can be
858 a bit noisy, and it doesn't always indicate a bug in the
859 program. However, it's generally considered good practice
860 to cover all the cases in your functions.</para>
865 <term><option>-fwarn-incomplete-record-updates</option>:</term>
867 <indexterm><primary><option>-fwarn-incomplete-record-updates</option></primary></indexterm>
868 <indexterm><primary>incomplete record updates, warning</primary></indexterm>
869 <indexterm><primary>record updates, incomplete</primary></indexterm>
872 <function>f</function> below will fail when applied to
873 <literal>Bar</literal>, so the compiler will emit a warning about
874 this when <option>-fwarn-incomplete-record-updates</option> is
878 data Foo = Foo { x :: Int }
882 f foo = foo { x = 6 }
885 <para>This option isn't enabled be default because it can be
886 very noisy, and it often doesn't indicate a bug in the
893 <option>-fwarn-misc</option>:
894 <indexterm><primary><option>-fwarn-misc</option></primary></indexterm>
897 <para>Turns on warnings for various harmless but untidy
898 things. This currently includes: importing a type with
899 <literal>(..)</literal> when the export is abstract, and
900 listing duplicate class assertions in a qualified type.</para>
906 <option>-fwarn-missing-fields</option>:
907 <indexterm><primary><option>-fwarn-missing-fields</option></primary></indexterm>
908 <indexterm><primary>missing fields, warning</primary></indexterm>
909 <indexterm><primary>fields, missing</primary></indexterm>
913 <para>This option is on by default, and warns you whenever
914 the construction of a labelled field constructor isn't
915 complete, missing initializers for one or more fields. While
916 not an error (the missing fields are initialised with
917 bottoms), it is often an indication of a programmer error.</para>
922 <term><option>-fwarn-missing-methods</option>:</term>
924 <indexterm><primary><option>-fwarn-missing-methods</option></primary></indexterm>
925 <indexterm><primary>missing methods, warning</primary></indexterm>
926 <indexterm><primary>methods, missing</primary></indexterm>
928 <para>This option is on by default, and warns you whenever
929 an instance declaration is missing one or more methods, and
930 the corresponding class declaration has no default
931 declaration for them.</para>
932 <para>The warning is suppressed if the method name
933 begins with an underscore. Here's an example where this is useful:
936 _simpleFn :: a -> String
937 complexFn :: a -> a -> String
938 complexFn x y = ... _simpleFn ...
940 The idea is that: (a) users of the class will only call <literal>complexFn</literal>;
941 never <literal>_simpleFn</literal>; and (b)
942 instance declarations can define either <literal>complexFn</literal> or <literal>_simpleFn</literal>.
948 <term><option>-fwarn-missing-signatures</option>:</term>
950 <indexterm><primary><option>-fwarn-missing-signatures</option></primary></indexterm>
951 <indexterm><primary>type signatures, missing</primary></indexterm>
953 <para>If you would like GHC to check that every top-level
954 function/value has a type signature, use the
955 <option>-fwarn-missing-signatures</option> option. This
956 option is off by default.</para>
961 <term><option>-fwarn-name-shadowing</option>:</term>
963 <indexterm><primary><option>-fwarn-name-shadowing</option></primary></indexterm>
964 <indexterm><primary>shadowing, warning</primary></indexterm>
966 <para>This option causes a warning to be emitted whenever an
967 inner-scope value has the same name as an outer-scope value,
968 i.e. the inner value shadows the outer one. This can catch
969 typographical errors that turn into hard-to-find bugs, e.g.,
970 in the inadvertent cyclic definition <literal>let x = ... x
971 ... in</literal>.</para>
973 <para>Consequently, this option does
974 <emphasis>will</emphasis> complain about cyclic recursive
980 <term><option>-fwarn-orphans</option>:</term>
982 <indexterm><primary><option>-fwarn-orphans</option></primary></indexterm>
983 <indexterm><primary>orphan instances, warning</primary></indexterm>
984 <indexterm><primary>orphan rules, warning</primary></indexterm>
986 <para>This option causes a warning to be emitted whenever the
987 module contains an "orphan" instance declaration or rewrite rule.
988 An instance declartion is an orphan if it appears in a module in
989 which neither the class nor the type being instanced are declared
990 in the same module. A rule is an orphan if it is a rule for a
991 function declared in another module. A module containing any
992 orphans is called an orphan module.</para>
993 <para>The trouble with orphans is that GHC must pro-actively read the interface
994 files for all orphan modules, just in case their instances or rules
995 play a role, whether or not the module's interface would otherwise
996 be of any use. Other things being equal, avoid orphan modules.</para>
1002 <option>-fwarn-overlapping-patterns</option>:
1003 <indexterm><primary><option>-fwarn-overlapping-patterns</option></primary></indexterm>
1004 <indexterm><primary>overlapping patterns, warning</primary></indexterm>
1005 <indexterm><primary>patterns, overlapping</primary></indexterm>
1008 <para>By default, the compiler will warn you if a set of
1009 patterns are overlapping, i.e.,</para>
1012 f :: String -> Int
1018 <para>where the last pattern match in <function>f</function>
1019 won't ever be reached, as the second pattern overlaps
1020 it. More often than not, redundant patterns is a programmer
1021 mistake/error, so this option is enabled by default.</para>
1026 <term><option>-fwarn-simple-patterns</option>:</term>
1028 <indexterm><primary><option>-fwarn-simple-patterns</option></primary>
1030 <para>Causes the compiler to warn about lambda-bound
1031 patterns that can fail, eg. <literal>\(x:xs)->...</literal>.
1032 Normally, these aren't treated as incomplete patterns by
1033 <option>-fwarn-incomplete-patterns</option>.</para>
1034 <para>``Lambda-bound patterns'' includes all places where there is a single pattern,
1035 including list comprehensions and do-notation. In these cases, a pattern-match
1036 failure is quite legitimate, and triggers filtering (list comprehensions) or
1037 the monad <literal>fail</literal> operation (monads). For example:
1039 f :: [Maybe a] -> [a]
1040 f xs = [y | Just y <- xs]
1042 Switching on <option>-fwarn-simple-patterns</option> will elicit warnings about
1043 these probably-innocent cases, which is why the flag is off by default. </para>
1044 <para> The <literal>deriving( Read )</literal> mechanism produces monadic code with
1045 pattern matches, so you will also get misleading warnings about the compiler-generated
1046 code. (This is arguably a Bad Thing, but it's awkward to fix.)</para>
1052 <term><option>-fwarn-type-defaults</option>:</term>
1054 <indexterm><primary><option>-fwarn-type-defaults</option></primary></indexterm>
1055 <indexterm><primary>defaulting mechanism, warning</primary></indexterm>
1056 <para>Have the compiler warn/inform you where in your source
1057 the Haskell defaulting mechanism for numeric types kicks
1058 in. This is useful information when converting code from a
1059 context that assumed one default into one with another,
1060 e.g., the `default default' for Haskell 1.4 caused the
1061 otherwise unconstrained value <constant>1</constant> to be
1062 given the type <literal>Int</literal>, whereas Haskell 98
1063 defaults it to <literal>Integer</literal>. This may lead to
1064 differences in performance and behaviour, hence the
1065 usefulness of being non-silent about this.</para>
1067 <para>This warning is off by default.</para>
1072 <term><option>-fwarn-unused-binds</option>:</term>
1074 <indexterm><primary><option>-fwarn-unused-binds</option></primary></indexterm>
1075 <indexterm><primary>unused binds, warning</primary></indexterm>
1076 <indexterm><primary>binds, unused</primary></indexterm>
1077 <para>Report any function definitions (and local bindings)
1078 which are unused. For top-level functions, the warning is
1079 only given if the binding is not exported.</para>
1080 <para>A definition is regarded as "used" if (a) it is exported, or (b) it is
1081 mentioned in the right hand side of another definition that is used, or (c) the
1082 function it defines begins with an underscore. The last case provides a
1083 way to suppress unused-binding warnings selectively. </para>
1084 <para> Notice that a variable
1085 is reported as unused even if it appears in the right-hand side of another
1086 unused binding. </para>
1091 <term><option>-fwarn-unused-imports</option>:</term>
1093 <indexterm><primary><option>-fwarn-unused-imports</option></primary></indexterm>
1094 <indexterm><primary>unused imports, warning</primary></indexterm>
1095 <indexterm><primary>imports, unused</primary></indexterm>
1097 <para>Report any modules that are explicitly imported but
1098 never used. However, the form <literal>import M()</literal> is
1099 never reported as an unused import, because it is a useful idiom
1100 for importing instance declarations, which are anonymous in Haskell.</para>
1105 <term><option>-fwarn-unused-matches</option>:</term>
1107 <indexterm><primary><option>-fwarn-unused-matches</option></primary></indexterm>
1108 <indexterm><primary>unused matches, warning</primary></indexterm>
1109 <indexterm><primary>matches, unused</primary></indexterm>
1111 <para>Report all unused variables which arise from pattern
1112 matches, including patterns consisting of a single variable.
1113 For instance <literal>f x y = []</literal> would report
1114 <varname>x</varname> and <varname>y</varname> as unused. The
1115 warning is suppressed if the variable name begins with an underscore, thus:
1125 <para>If you're feeling really paranoid, the
1126 <option>-dcore-lint</option>
1127 option<indexterm><primary><option>-dcore-lint</option></primary></indexterm>
1128 is a good choice. It turns on heavyweight intra-pass
1129 sanity-checking within GHC. (It checks GHC's sanity, not
1136 <sect1 id="options-optimise">
1137 <title>Optimisation (code improvement)</title>
1139 <indexterm><primary>optimisation</primary></indexterm>
1140 <indexterm><primary>improvement, code</primary></indexterm>
1142 <para>The <option>-O*</option> options specify convenient
1143 “packages” of optimisation flags; the
1144 <option>-f*</option> options described later on specify
1145 <emphasis>individual</emphasis> optimisations to be turned on/off;
1146 the <option>-m*</option> options specify
1147 <emphasis>machine-specific</emphasis> optimisations to be turned
1150 <sect2 id="optimise-pkgs">
1151 <title><option>-O*</option>: convenient “packages” of optimisation flags.</title>
1153 <para>There are <emphasis>many</emphasis> options that affect
1154 the quality of code produced by GHC. Most people only have a
1155 general goal, something like “Compile quickly” or
1156 “Make my program run like greased lightning.” The
1157 following “packages” of optimisations (or lack
1158 thereof) should suffice.</para>
1160 <para>Note that higher optimisation levels cause more
1161 cross-module optimisation to be performed, which can have an
1162 impact on how much of your program needs to be recompiled when
1163 you change something. This is one reaosn to stick to
1164 no-optimisation when developing code.</para>
1170 No <option>-O*</option>-type option specified:
1171 <indexterm><primary>-O* not specified</primary></indexterm>
1174 <para>This is taken to mean: “Please compile
1175 quickly; I'm not over-bothered about compiled-code
1176 quality.” So, for example: <command>ghc -c
1177 Foo.hs</command></para>
1183 <option>-O0</option>:
1184 <indexterm><primary><option>-O0</option></primary></indexterm>
1187 <para>Means “turn off all optimisation”,
1188 reverting to the same settings as if no
1189 <option>-O</option> options had been specified. Saying
1190 <option>-O0</option> can be useful if
1191 eg. <command>make</command> has inserted a
1192 <option>-O</option> on the command line already.</para>
1198 <option>-O</option> or <option>-O1</option>:
1199 <indexterm><primary>-O option</primary></indexterm>
1200 <indexterm><primary>-O1 option</primary></indexterm>
1201 <indexterm><primary>optimise</primary><secondary>normally</secondary></indexterm>
1204 <para>Means: “Generate good-quality code without
1205 taking too long about it.” Thus, for example:
1206 <command>ghc -c -O Main.lhs</command></para>
1208 <para><option>-O</option> currently also implies
1209 <option>-fvia-C</option>. This may change in the
1216 <option>-O2</option>:
1217 <indexterm><primary>-O2 option</primary></indexterm>
1218 <indexterm><primary>optimise</primary><secondary>aggressively</secondary></indexterm>
1221 <para>Means: “Apply every non-dangerous
1222 optimisation, even if it means significantly longer
1223 compile times.”</para>
1225 <para>The avoided “dangerous” optimisations
1226 are those that can make runtime or space
1227 <emphasis>worse</emphasis> if you're unlucky. They are
1228 normally turned on or off individually.</para>
1230 <para>At the moment, <option>-O2</option> is
1231 <emphasis>unlikely</emphasis> to produce better code than
1232 <option>-O</option>.</para>
1238 <option>-Ofile <file></option>:
1239 <indexterm><primary>-Ofile <file> option</primary></indexterm>
1240 <indexterm><primary>optimising, customised</primary></indexterm>
1243 <para>(NOTE: not supported since GHC 4.x. Please ask if
1244 you're interested in this.)</para>
1246 <para>For those who need <emphasis>absolute</emphasis>
1247 control over <emphasis>exactly</emphasis> what options are
1248 used (e.g., compiler writers, sometimes :-), a list of
1249 options can be put in a file and then slurped in with
1250 <option>-Ofile</option>.</para>
1252 <para>In that file, comments are of the
1253 <literal>#</literal>-to-end-of-line variety; blank
1254 lines and most whitespace is ignored.</para>
1256 <para>Please ask if you are baffled and would like an
1257 example of <option>-Ofile</option>!</para>
1262 <para>We don't use a <option>-O*</option> flag for day-to-day
1263 work. We use <option>-O</option> to get respectable speed;
1264 e.g., when we want to measure something. When we want to go for
1265 broke, we tend to use <option>-O2 -fvia-C</option> (and we go for
1266 lots of coffee breaks).</para>
1268 <para>The easiest way to see what <option>-O</option> (etc.)
1269 “really mean” is to run with <option>-v</option>,
1270 then stand back in amazement.</para>
1273 <sect2 id="options-f">
1274 <title><option>-f*</option>: platform-independent flags</title>
1276 <indexterm><primary>-f* options (GHC)</primary></indexterm>
1277 <indexterm><primary>-fno-* options (GHC)</primary></indexterm>
1279 <para>These flags turn on and off individual optimisations.
1280 They are normally set via the <option>-O</option> options
1281 described above, and as such, you shouldn't need to set any of
1282 them explicitly (indeed, doing so could lead to unexpected
1283 results). However, there are one or two that may be of
1288 <term><option>-fexcess-precision</option>:</term>
1290 <indexterm><primary><option>-fexcess-precision</option></primary></indexterm>
1291 <para>When this option is given, intermediate floating
1292 point values can have a <emphasis>greater</emphasis>
1293 precision/range than the final type. Generally this is a
1294 good thing, but some programs may rely on the exact
1296 <literal>Float</literal>/<literal>Double</literal> values
1297 and should not use this option for their compilation.</para>
1302 <term><option>-fignore-asserts</option>:</term>
1304 <indexterm><primary><option>-fignore-asserts</option></primary></indexterm>
1305 <para>Causes GHC to ignore uses of the function
1306 <literal>Exception.assert</literal> in source code (in
1307 other words, rewriting <literal>Exception.assert p
1308 e</literal> to <literal>e</literal> (see <xref
1309 linkend="sec-assertions"/>). This flag is turned on by
1310 <option>-O</option>.
1317 <option>-fno-cse</option>
1318 <indexterm><primary><option>-fno-cse</option></primary></indexterm>
1321 <para>Turns off the common-sub-expression elimination optimisation.
1322 Can be useful if you have some <literal>unsafePerformIO</literal>
1323 expressions that you don't want commoned-up.</para>
1329 <option>-fno-strictness</option>
1330 <indexterm><primary><option>-fno-strictness</option></primary></indexterm>
1333 <para>Turns off the strictness analyser; sometimes it eats
1334 too many cycles.</para>
1340 <option>-fno-full-laziness</option>
1341 <indexterm><primary><option>-fno-full-laziness</option></primary></indexterm>
1344 <para>Turns off the full laziness optimisation (also known as
1345 let-floating). Full laziness increases sharing, which can lead
1346 to increased memory residency.</para>
1352 <option>-fno-state-hack</option>
1353 <indexterm><primary><option>-fno-state-hack</option></primary></indexterm>
1356 <para>Turn off the "state hack" whereby any lambda with a
1357 <literal>State#</literal> token as argument is considered to be
1358 single-entry, hence it is considered OK to inline things inside
1359 it. This can improve performance of IO and ST monad code, but it
1360 runs the risk of reducing sharing.</para>
1366 <option>-funbox-strict-fields</option>:
1367 <indexterm><primary><option>-funbox-strict-fields</option></primary></indexterm>
1368 <indexterm><primary>strict constructor fields</primary></indexterm>
1369 <indexterm><primary>constructor fields, strict</primary></indexterm>
1372 <para>This option causes all constructor fields which are
1373 marked strict (i.e. “!”) to be unboxed or
1374 unpacked if possible. It is equivalent to adding an
1375 <literal>UNPACK</literal> pragma to every strict
1376 constructor field (see <xref
1377 linkend="unpack-pragma"/>).</para>
1379 <para>This option is a bit of a sledgehammer: it might
1380 sometimes make things worse. Selectively unboxing fields
1381 by using <literal>UNPACK</literal> pragmas might be
1388 <option>-funfolding-update-in-place<n></option>
1389 <indexterm><primary><option>-funfolding-update-in-place</option></primary></indexterm>
1392 <para>Switches on an experimental "optimisation".
1393 Switching it on makes the compiler a little keener to
1394 inline a function that returns a constructor, if the
1395 context is that of a thunk.
1399 If we inlined plusInt we might get an opportunity to use
1400 update-in-place for the thunk 'x'.</para>
1406 <option>-funfolding-creation-threshold<n></option>:
1407 <indexterm><primary><option>-funfolding-creation-threshold</option></primary></indexterm>
1408 <indexterm><primary>inlining, controlling</primary></indexterm>
1409 <indexterm><primary>unfolding, controlling</primary></indexterm>
1412 <para>(Default: 45) Governs the maximum size that GHC will
1413 allow a function unfolding to be. (An unfolding has a
1414 “size” that reflects the cost in terms of
1415 “code bloat” of expanding that unfolding at
1416 at a call site. A bigger function would be assigned a
1417 bigger cost.) </para>
1419 <para> Consequences: (a) nothing larger than this will be
1420 inlined (unless it has an INLINE pragma); (b) nothing
1421 larger than this will be spewed into an interface
1425 <para> Increasing this figure is more likely to result in longer
1426 compile times than faster code. The next option is more
1432 <term><option>-funfolding-use-threshold<n></option>:</term>
1434 <indexterm><primary><option>-funfolding-use-threshold</option></primary></indexterm>
1435 <indexterm><primary>inlining, controlling</primary></indexterm>
1436 <indexterm><primary>unfolding, controlling</primary></indexterm>
1438 <para>(Default: 8) This is the magic cut-off figure for
1439 unfolding: below this size, a function definition will be
1440 unfolded at the call-site, any bigger and it won't. The
1441 size computed for a function depends on two things: the
1442 actual size of the expression minus any discounts that
1443 apply (see <option>-funfolding-con-discount</option>).</para>
1454 <sect1 id="sec-using-concurrent">
1455 <title>Using Concurrent Haskell</title>
1457 <indexterm><primary>Concurrent Haskell—use</primary></indexterm>
1460 GHC supports Concurrent Haskell by default, without requiring a
1461 special option or libraries compiled in a certain way. To get access
1462 to the support libraries for Concurrent Haskell, just import
1463 <literal>Control.Concurrent</literal> (details are in the accompanying
1464 library documentation).</para>
1467 RTS options are provided for modifying the behaviour of the threaded
1468 runtime system. See <xref linkend="parallel-rts-opts"/>.
1472 Concurrent Haskell is described in more detail in the documentation
1473 for the <literal>Control.Concurrent</literal> module.
1478 <sect1 id="sec-using-parallel">
1479 <title>Using parallel Haskell</title>
1482 <indexterm><primary>parallel Haskell—use</primary></indexterm>
1486 [You won't be able to execute parallel Haskell programs unless PVM3
1487 (parallel Virtual Machine, version 3) is installed at your site.]
1491 To compile a Haskell program for parallel execution under PVM, use the
1492 <option>-parallel</option> option,<indexterm><primary>-parallel
1493 option</primary></indexterm> both when compiling <emphasis>and
1494 linking</emphasis>. You will probably want to <literal>import
1495 parallel</literal> into your Haskell modules.
1499 To run your parallel program, once PVM is going, just invoke it
1500 “as normal”. The main extra RTS option is
1501 <option>-qp<n></option>, to say how many PVM
1502 “processors” your program to run on. (For more details of
1503 all relevant RTS options, please see <xref
1504 linkend="parallel-rts-opts"/>.)
1508 In truth, running parallel Haskell programs and getting information
1509 out of them (e.g., parallelism profiles) is a battle with the vagaries of
1510 PVM, detailed in the following sections.
1513 <sect2 id="pvm-dummies">
1514 <title>Dummy's guide to using PVM</title>
1517 <indexterm><primary>PVM, how to use</primary></indexterm>
1518 <indexterm><primary>parallel Haskell—PVM use</primary></indexterm>
1519 Before you can run a parallel program under PVM, you must set the
1520 required environment variables (PVM's idea, not ours); something like,
1521 probably in your <filename>.cshrc</filename> or equivalent:
1524 setenv PVM_ROOT /wherever/you/put/it
1525 setenv PVM_ARCH `$PVM_ROOT/lib/pvmgetarch`
1526 setenv PVM_DPATH $PVM_ROOT/lib/pvmd
1532 Creating and/or controlling your “parallel machine” is a purely-PVM
1533 business; nothing specific to parallel Haskell. The following paragraphs
1534 describe how to configure your parallel machine interactively.
1538 If you use parallel Haskell regularly on the same machine configuration it
1539 is a good idea to maintain a file with all machine names and to make the
1540 environment variable PVM_HOST_FILE point to this file. Then you can avoid
1541 the interactive operations described below by just saying
1549 You use the <command>pvm</command><indexterm><primary>pvm command</primary></indexterm> command to start PVM on your
1550 machine. You can then do various things to control/monitor your
1551 “parallel machine;” the most useful being:
1557 <colspec align="left"/>
1561 <entry><keycombo><keycap>Control</keycap><keycap>D</keycap></keycombo></entry>
1562 <entry>exit <command>pvm</command>, leaving it running</entry>
1566 <entry><command>halt</command></entry>
1567 <entry>kill off this “parallel machine” & exit</entry>
1571 <entry><command>add <host></command></entry>
1572 <entry>add <command><host></command> as a processor</entry>
1576 <entry><command>delete <host></command></entry>
1577 <entry>delete <command><host></command></entry>
1581 <entry><command>reset</command></entry>
1582 <entry>kill what's going, but leave PVM up</entry>
1586 <entry><command>conf</command></entry>
1587 <entry>list the current configuration</entry>
1591 <entry><command>ps</command></entry>
1592 <entry>report processes' status</entry>
1596 <entry><command>pstat <pid></command></entry>
1597 <entry>status of a particular process</entry>
1606 The PVM documentation can tell you much, much more about <command>pvm</command>!
1611 <sect2 id="par-profiles">
1612 <title>parallelism profiles</title>
1615 <indexterm><primary>parallelism profiles</primary></indexterm>
1616 <indexterm><primary>profiles, parallelism</primary></indexterm>
1617 <indexterm><primary>visualisation tools</primary></indexterm>
1621 With parallel Haskell programs, we usually don't care about the
1622 results—only with “how parallel” it was! We want pretty pictures.
1626 parallelism profiles (à la <command>hbcpp</command>) can be generated with the
1627 <option>-qP</option><indexterm><primary>-qP RTS option (concurrent, parallel)</primary></indexterm> RTS option. The
1628 per-processor profiling info is dumped into files named
1629 <filename><full-path><program>.gr</filename>. These are then munged into a PostScript picture,
1630 which you can then display. For example, to run your program
1631 <filename>a.out</filename> on 8 processors, then view the parallelism profile, do:
1637 <prompt>$</prompt> ./a.out +RTS -qP -qp8
1638 <prompt>$</prompt> grs2gr *.???.gr > temp.gr # combine the 8 .gr files into one
1639 <prompt>$</prompt> gr2ps -O temp.gr # cvt to .ps; output in temp.ps
1640 <prompt>$</prompt> ghostview -seascape temp.ps # look at it!
1646 The scripts for processing the parallelism profiles are distributed
1647 in <filename>ghc/utils/parallel/</filename>.
1653 <title>Other useful info about running parallel programs</title>
1656 The “garbage-collection statistics” RTS options can be useful for
1657 seeing what parallel programs are doing. If you do either
1658 <option>+RTS -Sstderr</option><indexterm><primary>-Sstderr RTS option</primary></indexterm> or <option>+RTS -sstderr</option>, then
1659 you'll get mutator, garbage-collection, etc., times on standard
1660 error. The standard error of all PE's other than the `main thread'
1661 appears in <filename>/tmp/pvml.nnn</filename>, courtesy of PVM.
1665 Whether doing <option>+RTS -Sstderr</option> or not, a handy way to watch
1666 what's happening overall is: <command>tail -f /tmp/pvml.nnn</command>.
1671 <sect2 id="parallel-rts-opts">
1672 <title>RTS options for Concurrent/parallel Haskell
1676 <indexterm><primary>RTS options, concurrent</primary></indexterm>
1677 <indexterm><primary>RTS options, parallel</primary></indexterm>
1678 <indexterm><primary>Concurrent Haskell—RTS options</primary></indexterm>
1679 <indexterm><primary>parallel Haskell—RTS options</primary></indexterm>
1683 Besides the usual runtime system (RTS) options
1684 (<xref linkend="runtime-control"/>), there are a few options particularly
1685 for concurrent/parallel execution.
1692 <term><option>-qp<N></option>:</term>
1695 <indexterm><primary>-qp<N> RTS option</primary></indexterm>
1696 (paraLLEL ONLY) Use <literal><N></literal> PVM processors to run this program;
1702 <term><option>-C[<us>]</option>:</term>
1705 <indexterm><primary>-C<us> RTS option</primary></indexterm> Sets
1706 the context switch interval to <literal><s></literal> seconds.
1707 A context switch will occur at the next heap block allocation after
1708 the timer expires (a heap block allocation occurs every 4k of
1709 allocation). With <option>-C0</option> or <option>-C</option>,
1710 context switches will occur as often as possible (at every heap block
1711 allocation). By default, context switches occur every 20ms
1712 milliseconds. Note that GHC's internal timer ticks every 20ms, and
1713 the context switch timer is always a multiple of this timer, so 20ms
1714 is the maximum granularity available for timed context switches.
1719 <term><option>-q[v]</option>:</term>
1722 <indexterm><primary>-q RTS option</primary></indexterm>
1723 (paraLLEL ONLY) Produce a quasi-parallel profile of thread activity,
1724 in the file <filename><program>.qp</filename>. In the style of <command>hbcpp</command>, this profile
1725 records the movement of threads between the green (runnable) and red
1726 (blocked) queues. If you specify the verbose suboption (<option>-qv</option>), the
1727 green queue is split into green (for the currently running thread
1728 only) and amber (for other runnable threads). We do not recommend
1729 that you use the verbose suboption if you are planning to use the
1730 <command>hbcpp</command> profiling tools or if you are context switching at every heap
1731 check (with <option>-C</option>).
1737 <term><option>-qt<num></option>:</term>
1740 <indexterm><primary>-qt<num> RTS option</primary></indexterm>
1741 (paraLLEL ONLY) Limit the thread pool size, i.e. the number of concurrent
1742 threads per processor to <literal><num></literal>. The default is
1743 32. Each thread requires slightly over 1K <emphasis>words</emphasis> in
1744 the heap for thread state and stack objects. (For 32-bit machines, this
1745 translates to 4K bytes, and for 64-bit machines, 8K bytes.)
1751 <term><option>-d</option>:</term>
1754 <indexterm><primary>-d RTS option (parallel)</primary></indexterm>
1755 (paraLLEL ONLY) Turn on debugging. It pops up one xterm (or GDB, or
1756 something…) per PVM processor. We use the standard <command>debugger</command>
1757 script that comes with PVM3, but we sometimes meddle with the
1758 <command>debugger2</command> script. We include ours in the GHC distribution,
1759 in <filename>ghc/utils/pvm/</filename>.
1765 <term><option>-qe<num></option>:</term>
1768 <indexterm><primary>-qe<num> RTS option
1769 (parallel)</primary></indexterm> (paraLLEL ONLY) Limit the spark pool size
1770 i.e. the number of pending sparks per processor to
1771 <literal><num></literal>. The default is 100. A larger number may be
1772 appropriate if your program generates large amounts of parallelism
1778 <term><option>-qQ<num></option>:</term>
1781 <indexterm><primary>-qQ<num> RTS option (parallel)</primary></indexterm>
1782 (paraLLEL ONLY) Set the size of packets transmitted between processors
1783 to <literal><num></literal>. The default is 1024 words. A larger number may be
1784 appropriate if your machine has a high communication cost relative to
1790 <term><option>-qh<num></option>:</term>
1793 <indexterm><primary>-qh<num> RTS option (parallel)</primary></indexterm>
1794 (paraLLEL ONLY) Select a packing scheme. Set the number of non-root thunks to pack in one packet to
1795 <num>-1 (0 means infinity). By default GUM uses full-subgraph
1796 packing, i.e. the entire subgraph with the requested closure as root is
1797 transmitted (provided it fits into one packet). Choosing a smaller value
1798 reduces the amount of pre-fetching of work done in GUM. This can be
1799 advantageous for improving data locality but it can also worsen the balance
1800 of the load in the system.
1805 <term><option>-qg<num></option>:</term>
1808 <indexterm><primary>-qg<num> RTS option
1809 (parallel)</primary></indexterm> (paraLLEL ONLY) Select a globalisation
1810 scheme. This option affects the
1811 generation of global addresses when transferring data. Global addresses are
1812 globally unique identifiers required to maintain sharing in the distributed
1813 graph structure. Currently this is a binary option. With <num>=0 full globalisation is used
1814 (default). This means a global address is generated for every closure that
1815 is transmitted. With <num>=1 a thunk-only globalisation scheme is
1816 used, which generated global address only for thunks. The latter case may
1817 lose sharing of data but has a reduced overhead in packing graph structures
1818 and maintaining internal tables of global addresses.
1829 <sect1 id="options-platform">
1830 <title>Platform-specific Flags</title>
1832 <indexterm><primary>-m* options</primary></indexterm>
1833 <indexterm><primary>platform-specific options</primary></indexterm>
1834 <indexterm><primary>machine-specific options</primary></indexterm>
1836 <para>Some flags only make sense for particular target
1842 <term><option>-mv8</option>:</term>
1844 <para>(SPARC machines)<indexterm><primary>-mv8 option (SPARC
1845 only)</primary></indexterm> Means to pass the like-named
1846 option to GCC; it says to use the Version 8 SPARC
1847 instructions, notably integer multiply and divide. The
1848 similar <option>-m*</option> GCC options for SPARC also
1849 work, actually.</para>
1854 <term><option>-monly-[32]-regs</option>:</term>
1856 <para>(iX86 machines)<indexterm><primary>-monly-N-regs
1857 option (iX86 only)</primary></indexterm> GHC tries to
1858 “steal” four registers from GCC, for performance
1859 reasons; it almost always works. However, when GCC is
1860 compiling some modules with four stolen registers, it will
1861 crash, probably saying:
1864 Foo.hc:533: fixed or forbidden register was spilled.
1865 This may be due to a compiler bug or to impossible asm
1866 statements or clauses.
1869 Just give some registers back with
1870 <option>-monly-N-regs</option>. Try `3' first, then `2'.
1871 If `2' doesn't work, please report the bug to us.</para>
1880 <sect1 id="ext-core">
1881 <title>Generating and compiling External Core Files</title>
1883 <indexterm><primary>intermediate code generation</primary></indexterm>
1885 <para>GHC can dump its optimized intermediate code (said to be in “Core” format)
1886 to a file as a side-effect of compilation. Core files, which are given the suffix
1887 <filename>.hcr</filename>, can be read and processed by non-GHC back-end
1888 tools. The Core format is formally described in <ulink url="http://www.haskell.org/ghc/docs/papers/core.ps.gz">
1889 <citetitle>An External Representation for the GHC Core Language</citetitle></ulink>,
1890 and sample tools (in Haskell)
1891 for manipulating Core files are available in the GHC source distribution
1892 directory <literal>/fptools/ghc/utils/ext-core</literal>.
1893 Note that the format of <literal>.hcr</literal>
1894 files is <emphasis>different</emphasis> (though similar) to the Core output format generated
1895 for debugging purposes (<xref linkend="options-debugging"/>).</para>
1897 <para>The Core format natively supports notes which you can add to
1898 your source code using the <literal>CORE</literal> pragma (see <xref
1899 linkend="pragmas"/>).</para>
1905 <option>-fext-core</option>
1906 <indexterm><primary><option>-fext-core</option></primary></indexterm>
1909 <para>Generate <literal>.hcr</literal> files.</para>
1915 <para>GHC can also read in External Core files as source; just give the <literal>.hcr</literal> file on
1916 the command line, instead of the <literal>.hs</literal> or <literal>.lhs</literal> Haskell source.
1917 A current infelicity is that you need to give the <literal>-fglasgow-exts</literal> flag too, because
1918 ordinary Haskell 98, when translated to External Core, uses things like rank-2 types.</para>
1927 ;;; Local Variables: ***
1929 ;;; sgml-parent-document: ("users_guide.xml" "book" "chapter") ***