1 <chapter id="using-ghc">
2 <title>Using GHC</title>
4 <indexterm><primary>GHC, using</primary></indexterm>
5 <indexterm><primary>using GHC</primary></indexterm>
8 <title>Options overview</title>
10 <para>GHC's behaviour is controlled by
11 <firstterm>options</firstterm>, which for historical reasons are
12 also sometimes referred to as command-line flags or arguments.
13 Options can be specified in three ways:</para>
16 <title>Command-line arguments</title>
18 <indexterm><primary>structure, command-line</primary></indexterm>
19 <indexterm><primary>command-line</primary><secondary>arguments</secondary></indexterm>
20 <indexterm><primary>arguments</primary><secondary>command-line</secondary></indexterm>
22 <para>An invocation of GHC takes the following form:</para>
28 <para>Command-line arguments are either options or file names.</para>
30 <para>Command-line options begin with <literal>-</literal>.
31 They may <emphasis>not</emphasis> be grouped:
32 <option>-vO</option> is different from <option>-v -O</option>.
33 Options need not precede filenames: e.g., <literal>ghc *.o -o
34 foo</literal>. All options are processed and then applied to
35 all files; you cannot, for example, invoke <literal>ghc -c -O1
36 Foo.hs -O2 Bar.hs</literal> to apply different optimisation
37 levels to the files <filename>Foo.hs</filename> and
38 <filename>Bar.hs</filename>.</para>
41 <Sect2 id="source-file-options">
42 <title>Command line options in source files</title>
44 <indexterm><primary>source-file options</primary></indexterm>
46 <para>Sometimes it is useful to make the connection between a
47 source file and the command-line options it requires quite
48 tight. For instance, if a Haskell source file uses GHC
49 extensions, it will always need to be compiled with the
50 <option>-fglasgow-exts</option> option. Rather than maintaining
51 the list of per-file options in a <filename>Makefile</filename>,
52 it is possible to do this directly in the source file using the
53 <literal>OPTIONS</literal> pragma <indexterm><primary>OPTIONS
54 pragma</primary></indexterm>:</para>
57 {-# OPTIONS -fglasgow-exts #-}
62 <para><literal>OPTIONS</literal> pragmas are only looked for at
63 the top of your source files, upto the first
64 (non-literate,non-empty) line not containing
65 <literal>OPTIONS</literal>. Multiple <literal>OPTIONS</literal>
66 pragmas are recognised. Note that your command shell does not
67 get to the source file options, they are just included literally
68 in the array of command-line arguments the compiler driver
69 maintains internally, so you'll be desperately disappointed if
70 you try to glob etc. inside <literal>OPTIONS</literal>.</para>
72 <para>NOTE: the contents of OPTIONS are prepended to the
73 command-line options, so you <emphasis>do</emphasis> have the
74 ability to override OPTIONS settings via the command
77 <para>It is not recommended to move all the contents of your
78 Makefiles into your source files, but in some circumstances, the
79 <literal>OPTIONS</literal> pragma is the Right Thing. (If you
80 use <option>-keep-hc-file-too</option> and have OPTION flags in
81 your module, the OPTIONS will get put into the generated .hc
86 <title>Setting options in GHCi</title>
88 <para>Options may also be modified from within GHCi, using the
89 <literal>:set</literal> command. See <xref linkend="ghci-set">
90 for more details.</para>
94 <sect1 id="static-dynamic-flags">
95 <title>Static vs. Dynamic options</title>
96 <indexterm><primary>static</primary><secondary>options</secondary>
98 <indexterm><primary>dynamic</primary><secondary>options</secondary>
101 <para>Each of GHC's command line options is classified as either
102 <firstterm>static</firstterm> or <firstterm>dynamic</firstterm>.
103 A static flag may only be specified on the command line, whereas a
104 dynamic flag may also be given in an <literal>OPTIONS</literal>
105 pragma in a source file or set from the GHCi command-line with
106 <literal>:set</literal>.</para>
108 <para>As a rule of thumb, all the language options are dynamic, as
109 are the warning options and the debugging options. The rest are
110 static, with the notable exceptions of <option>-v</option>,
111 <option>-cpp</option>, <option>-fasm</option>,
112 <option>-fvia-C</option>, and <option>-#include</option>.
114 The flag reference tables (<xref linkend="flag-reference">) lists
115 the status of each flag.</para>
118 <sect1 id="file-suffixes">
119 <title>Meaningful file suffixes</title>
121 <indexterm><primary>suffixes, file</primary></indexterm>
122 <indexterm><primary>file suffixes for GHC</primary></indexterm>
124 <para>File names with “meaningful” suffixes (e.g.,
125 <filename>.lhs</filename> or <filename>.o</filename>) cause the
126 “right thing” to happen to those files.</para>
131 <term><filename>.lhs</filename></term>
132 <indexterm><primary><literal>lhs</literal> suffix</primary></indexterm>
134 <para>A “literate Haskell” module.</para>
139 <term><filename>.hs</filename></term>
141 <para>A not-so-literate Haskell module.</para>
146 <term><filename>.hi</filename></term>
148 <para>A Haskell interface file, probably
149 compiler-generated.</para>
154 <term><filename>.hc</filename></term>
156 <para>Intermediate C file produced by the Haskell
162 <term><filename>.c</filename></term>
164 <para>A C file not produced by the Haskell
170 <term><filename>.s</filename></term>
172 <para>An assembly-language source file, usually produced by
178 <term><filename>.o</filename></term>
180 <para>An object file, produced by an assembler.</para>
185 <para>Files with other suffixes (or without suffixes) are passed
186 straight to the linker.</para>
191 <title>Modes of operation</title>
193 <para>GHC's behaviour is firstly controlled by a mode flag. Only
194 one of these flags may be given, but it does not necessarily need
195 to be the first option on the command-line. The available modes
200 <term><cmdsynopsis><command>ghc</command>
201 <arg choice=plain>––interactive</arg>
202 </cmdsynopsis></term>
203 <indexterm><primary>interactive mode</primary>
205 <indexterm><primary>ghci</primary>
208 <para>Interactive mode, which is also available as
209 <command>ghci</command>. Interactive mode is described in
210 more detail in <xref linkend="ghci">.</para>
215 <term><cmdsynopsis><command>ghc</command>
216 <arg choice=plain>––make</arg>
217 </cmdsynopsis></term>
218 <indexterm><primary>make mode</primary>
220 <indexterm><primary><option>––make</option></primary>
223 <para>In this mode, GHC will build a multi-module Haskell
224 program automatically, figuring out dependencies for itself.
225 If you have a straightforward Haskell program, this is
226 likely to be much easier, and faster, than using
227 <command>make</command>. Make mode is described in <xref
228 linkend="make-mode">.</para>
233 <term><cmdsynopsis><command>ghc</command>
234 <arg choice=plain>–e</arg> <arg choice=plain><replaceable>expr</replaceable></arg>
235 </cmdsynopsis></term>
236 <indexterm><primary>eval mode</primary>
239 <para>Expression-evaluation mode. This is very similar to
240 interactive mode, except that there is a single expression
241 to evaluate (<replaceable>expr</replaceable>) which is given
242 on the command line. See <xref linkend="eval-mode"> for
249 <command>ghc</command>
256 </cmdsynopsis></term>
257 <indexterm><primary><option>-E</option></primary></indexterm>
258 <indexterm><primary><option>-C</option></primary></indexterm>
259 <indexterm><primary><option>-S</option></primary></indexterm>
260 <indexterm><primary><option>-c</option></primary></indexterm>
262 <para>This is the traditional batch-compiler mode, in which
263 GHC can compile source files one at a time, or link objects
264 together into an executable. This mode also applies if
265 there is no other mode flag specified on the command line,
266 in which case it means that the specified files should be
267 compiled and then linked to form a program. See <xref
268 linkend="options-order">.</para>
273 <term><cmdsynopsis><command>ghc</command>
274 <arg choice=plain>–M</arg></cmdsynopsis></term>
275 <indexterm><primary>dependency-generation mode</primary>
278 <para>Dependency-generation mode. In this mode, GHC can be
279 used to generate dependency information suitable for use in
280 a <literal>Makefile</literal>. See <xref
281 linkend="sec-makefile-dependencies">.</para>
286 <term><cmdsynopsis><command>ghc</command>
287 <arg choice=plain>––mk-dll</arg></cmdsynopsis></term>
288 <indexterm><primary>dependency-generation mode</primary>
291 <para>DLL-creation mode (Windows only). See <xref
292 linkend="win32-dlls-create">.</para>
297 <sect2 id="make-mode">
298 <title>Using <command>ghc</command> <option>––make</option></title>
300 <indexterm><primary><option>––make</option></primary>
302 <indexterm><primary>separate compilation</primary>
305 <para>When given the <option>––make</option> option,
306 GHC will build a multi-module Haskell program by following
307 dependencies from a single root module (usually
308 <literal>Main</literal>). For example, if your
309 <literal>Main</literal> module is in a file called
310 <filename>Main.hs</filename>, you could compile and link the
311 program like this:</para>
314 ghc ––make Main.hs
317 <para>The command line may contain any number of source file
318 names or module names; GHC will figure out all the modules in
319 the program by following the imports from these initial modules.
320 It will then attempt to compile each module which is out of
321 date, and finally if there is a <literal>Main</literal> module,
322 the program will also be linked into an executable.</para>
324 <para>The main advantages to using <literal>ghc
325 ––make</literal> over traditional
326 <literal>Makefile</literal>s are:</para>
330 <para>GHC doesn't have to be restarted for each compilation,
331 which means it can cache information between compilations.
332 Compiling a muli-module program with <literal>ghc
333 ––make</literal> can be up to twice as fast as
334 running <literal>ghc</literal> individually on each source
338 <para>You don't have to write a
339 <literal>Makefile</literal>.</para>
341 <indexterm><primary><literal>Makefile</literal>s</primary><secondary>avoiding</secondary>
344 <para>GHC re-calculates the dependencies each time it is
345 invoked, so the dependencies never get out of sync with the
350 <para>Any of the command-line options described in the rest of
351 this chapter can be used with
352 <option>––make</option>, but note that any options
353 you give on the command line will apply to all the source files
354 compiled, so if you want any options to apply to a single source
355 file only, you'll need to use an <literal>OPTIONS</literal>
356 pragma (see <xref linkend="source-file-options">).</para>
358 <para>If the program needs to be linked with additional objects
359 (say, some auxilliary C code), then the object files can be
360 given on the command line and GHC will include them when linking
361 the executable.</para>
363 <para>Note that GHC can only follow dependencies if it has the
364 source file available, so if your program includes a module for
365 which there is no source file, even if you have an object and an
366 interface file for the module, then GHC will complain. The
367 exception to this rule is for package modules, which may or may
368 not have source files.</para>
370 <para>The source files for the program don't all need to be in
371 the same directory; the <option>-i</option> option can be used
372 to add directories to the search path (see <xref
373 linkend="search-path">).</para>
376 <sect2 id="eval-mode">
377 <title>Expression evaluation mode</title>
379 <para>This mode is very similar to interactive mode, except that
380 there is a single expression to evaluate which is specified on
381 the command line as an argument to the <option>-e</option>
385 ghc -e <replaceable>expr</replaceable>
388 <para>Haskell source files may be named on the command line, and
389 they will be loaded exactly as in interactive mode. The
390 expression is evaluated in the context of the loaded
393 <para>For example, to load and run a Haskell program containing
394 a module <literal>Main</literal>, we might say</para>
397 ghc -e Main.main Main.hs
400 <para>or we can just use this mode to evaluate expressions in
401 the context of the <literal>Prelude</literal>:</para>
404 $ ghc -e "interact (unlines.map reverse.lines)"
410 <sect2 id="options-order">
411 <title>Batch compiler mode</title>
413 <para>In this mode, GHC will compile one or more source files
414 given on the command line.</para>
416 <para>The first phase to run is determined by each input-file
417 suffix, and the last phase is determined by a flag. If no
418 relevant flag is present, then go all the way through linking.
419 This table summarises:</para>
423 <colspec align="left">
424 <colspec align="left">
425 <colspec align="left">
426 <colspec align="left">
430 <entry>Phase of the compilation system</entry>
431 <entry>Suffix saying “start here”</entry>
432 <entry>Flag saying “stop after”</entry>
433 <entry>(suffix of) output file</entry>
438 <entry>literate pre-processor</entry>
439 <entry><literal>.lhs</literal></entry>
441 <entry><literal>.hs</literal></entry>
445 <entry>C pre-processor (opt.) </entry>
446 <entry><literal>.hs</literal> (with
447 <option>-cpp</option>)</entry>
448 <entry><option>-E</option></entry>
449 <entry><literal>.hspp</literal></entry>
453 <entry>Haskell compiler</entry>
454 <entry><literal>.hs</literal></entry>
455 <entry><option>-C</option>, <option>-S</option></entry>
456 <entry><literal>.hc</literal>, <literal>.s</literal></entry>
460 <entry>C compiler (opt.)</entry>
461 <entry><literal>.hc</literal> or <literal>.c</literal></entry>
462 <entry><option>-S</option></entry>
463 <entry><literal>.s</literal></entry>
467 <entry>assembler</entry>
468 <entry><literal>.s</literal></entry>
469 <entry><option>-c</option></entry>
470 <entry><literal>.o</literal></entry>
474 <entry>linker</entry>
475 <entry><replaceable>other</replaceable></entry>
477 <entry><filename>a.out</filename></entry>
483 <indexterm><primary><option>-C</option></primary></indexterm>
484 <indexterm><primary><option>-E</option></primary></indexterm>
485 <indexterm><primary><option>-S</option></primary></indexterm>
486 <indexterm><primary><option>-c</option></primary></indexterm>
488 <para>Thus, a common invocation would be: </para>
491 ghc -c Foo.hs</screen>
493 <para>to compile the Haskell source file
494 <filename>Foo.hs</filename> to an object file
495 <filename>Foo.o</filename>.</para>
497 <para>Note: What the Haskell compiler proper produces depends on
498 whether a native-code generator<indexterm><primary>native-code
499 generator</primary></indexterm> is used (producing assembly
500 language) or not (producing C). See <xref
501 linkend="options-codegen"> for more details.</para>
503 <para>Note: C pre-processing is optional, the
504 <option>-cpp</option><indexterm><primary><option>-cpp</option></primary></indexterm>
505 flag turns it on. See <xref linkend="c-pre-processor"> for more
508 <para>Note: The option <option>-E</option><IndexTerm><Primary>-E
509 option</Primary></IndexTerm> runs just the pre-processing passes
510 of the compiler, dumping the result in a file. Note that this
511 differs from the previous behaviour of dumping the file to
512 standard output.</para>
516 <sect1 id="options-help">
517 <title>Help and verbosity options</title>
519 <IndexTerm><Primary>help options</Primary></IndexTerm>
520 <IndexTerm><Primary>verbosity options</Primary></IndexTerm>
524 <term><option>––help</option></term>
525 <term><option>-?</option></term>
526 <indexterm><primary><option>-?</option></primary></indexterm>
527 <indexterm><primary><option>––help</option></primary></indexterm>
529 <para>Cause GHC to spew a long usage message to standard
530 output and then exit.</para>
535 <term><option>-v</option></term>
536 <indexterm><primary><option>-v</option></primary></indexterm>
538 <para>The <option>-v</option> option makes GHC
539 <emphasis>verbose</emphasis>: it reports its version number
540 and shows (on stderr) exactly how it invokes each phase of
541 the compilation system. Moreover, it passes the
542 <option>-v</option> flag to most phases; each reports its
543 version number (and possibly some other information).</para>
545 <para>Please, oh please, use the <option>-v</option> option
546 when reporting bugs! Knowing that you ran the right bits in
547 the right order is always the first thing we want to
553 <term><option>-v</option><replaceable>n</replaceable></term>
554 <indexterm><primary><option>-v</option></primary></indexterm>
556 <para>To provide more control over the compiler's verbosity,
557 the <option>-v</option> flag takes an optional numeric
558 argument. Specifying <option>-v</option> on its own is
559 equivalent to <option>-v3</option>, and the other levels
560 have the following meanings:</para>
564 <term><option>-v0</option></term>
566 <para>Disable all non-essential messages (this is the
572 <term><option>-v1</option></term>
574 <para>Minimal verbosity: print one line per
575 compilation (this is the default when
576 <option>––make</option> or
577 <option>––interactive</option> is on).</para>
582 <term><option>-v2</option></term>
584 <para>Print the name of each compilation phase as it
585 is executed. (equivalent to
586 <option>-dshow-passes</option>).</para>
591 <term><option>-v3</option></term>
593 <para>The same as <option>-v2</option>, except that in
594 addition the full command line (if appropriate) for
595 each compilation phase is also printed.</para>
600 <term><option>-v4</option></term>
602 <para>The same as <option>-v3</option> except that the
603 intermediate program representation after each
604 compilation phase is also printed (excluding
605 preprocessed and C/assembly files).</para>
613 <term><option>-V</option></term>
614 <term><option>––version</option></term>
615 <indexterm><primary><option>-V</option></primary></indexterm>
616 <indexterm><primary><option>––version</option></primary></indexterm>
618 <para>Print a one-line string including GHC's version number.</para>
623 <term><option>––numeric-version</option></term>
624 <indexterm><primary><option>––numeric-version</option></primary></indexterm>
626 <para>Print GHC's numeric version number only.</para>
631 <term><option>––print-libdir</option></term>
632 <indexterm><primary><option>––print-libdir</option></primary></indexterm>
634 <para>Print the path to GHC's library directory. This is
635 the top of the directory tree containing GHC's libraries,
636 interfaces, and include files (usually something like
637 <literal>/usr/local/lib/ghc-5.04</literal> on Unix). This
639 <literal>$libdir</literal><indexterm><primary><literal>libdir</literal></primary>
640 </indexterm>in the package configuration file (see <xref
641 linkend="packages">).</para>
650 <sect1 id="options-sanity">
651 <title>Warnings and sanity-checking</title>
653 <indexterm><primary>sanity-checking options</primary></indexterm>
654 <indexterm><primary>warnings</primary></indexterm>
657 <para>GHC has a number of options that select which types of
658 non-fatal error messages, otherwise known as warnings, can be
659 generated during compilation. By default, you get a standard set
660 of warnings which are generally likely to indicate bugs in your
662 <option>-fwarn-overlapping-patterns</option>,
663 <option>-fwarn-deprecations</option>,
664 <option>-fwarn-duplicate-exports</option>,
665 <option>-fwarn-missing-fields</option>, and
666 <option>-fwarn-missing-methods</option>. The following flags are
667 simple ways to select standard “packages” of warnings:
673 <term><option>-W</option>:</term>
675 <IndexTerm><Primary>-W option</Primary></IndexTerm>
676 <para>Provides the standard warnings plus
677 <option>-fwarn-incomplete-patterns</option>,
678 <option>-fwarn-unused-matches</option>,
679 <option>-fwarn-unused-imports</option>,
680 <option>-fwarn-misc</option>, and
681 <option>-fwarn-unused-binds</option>.</para>
686 <term><option>-w</option>:</term>
688 <IndexTerm><Primary><option>-w</option></Primary></IndexTerm>
689 <para>Turns off all warnings, including the standard ones.</para>
694 <term><option>-Wall</option>:</term>
696 <indexterm><primary><option>-Wall</option></primary></indexterm>
697 <para>Turns on all warning options.</para>
702 <term><option>-Werror</option>:</term>
704 <indexterm><primary><option>-Werror</option></primary></indexterm>
705 <para>Makes any warning into a fatal error. Useful so that you don't
706 miss warnings when doing batch compilation. </para>
712 <para>The full set of warning options is described below. To turn
713 off any warning, simply give the corresponding
714 <option>-fno-warn-...</option> option on the command line.</para>
719 <term><option>-fwarn-deprecations</option>:</term>
721 <indexterm><primary><option>-fwarn-deprecations</option></primary>
723 <indexterm><primary>deprecations</primary></indexterm>
724 <para>Causes a warning to be emitted when a deprecated
725 function or type is used. Entities can be marked as
726 deprecated using a pragma, see <xref
727 linkend="deprecated-pragma">.</para>
732 <term><option>-fwarn-duplicate-exports</option>:</term>
734 <indexterm><primary><option>-fwarn-duplicate-exports</option></primary></indexterm>
735 <indexterm><primary>duplicate exports, warning</primary></indexterm>
736 <indexterm><primary>export lists, duplicates</primary></indexterm>
738 <para>Have the compiler warn about duplicate entries in
739 export lists. This is useful information if you maintain
740 large export lists, and want to avoid the continued export
741 of a definition after you've deleted (one) mention of it in
742 the export list.</para>
744 <para>This option is on by default.</para>
749 <term><option>-fwarn-hi-shadowing</option>:</term>
751 <indexterm><primary><option>-fwarn-hi-shadowing</option></primary></indexterm>
752 <indexterm><primary>shadowing</primary>
753 <secondary>interface files</secondary></indexterm>
755 <para>Causes the compiler to emit a warning when a module or
756 interface file in the current directory is shadowing one
757 with the same module name in a library or other
763 <term><option>-fwarn-incomplete-patterns</option>:</term>
765 <indexterm><primary><option>-fwarn-incomplete-patterns</option></primary></indexterm>
766 <indexterm><primary>incomplete patterns, warning</primary></indexterm>
767 <indexterm><primary>patterns, incomplete</primary></indexterm>
769 <para>Similarly for incomplete patterns, the function
770 <function>g</function> below will fail when applied to
771 non-empty lists, so the compiler will emit a warning about
772 this when <option>-fwarn-incomplete-patterns</option> is
779 <para>This option isn't enabled be default because it can be
780 a bit noisy, and it doesn't always indicate a bug in the
781 program. However, it's generally considered good practice
782 to cover all the cases in your functions.</para>
787 <term><option>-fwarn-misc</option>:</term>
788 <indexterm><primary><option>-fwarn-misc</option></primary></indexterm>
790 <para>Turns on warnings for various harmless but untidy
791 things. This currently includes: importing a type with
792 <literal>(..)</literal> when the export is abstract, and
793 listing duplicate class assertions in a qualified type.</para>
798 <term><option>-fwarn-missing-fields</option>:</term>
800 <indexterm><primary><option>-fwarn-missing-fields</option></primary></indexterm>
801 <indexterm><primary>missing fields, warning</primary></indexterm>
802 <indexterm><primary>fields, missing</primary></indexterm>
804 <para>This option is on by default, and warns you whenever
805 the construction of a labelled field constructor isn't
806 complete, missing initializers for one or more fields. While
807 not an error (the missing fields are initialised with
808 bottoms), it is often an indication of a programmer error.</para>
813 <term><option>-fwarn-missing-methods</option>:</term>
815 <indexterm><primary><option>-fwarn-missing-methods</option></primary></indexterm>
816 <indexterm><primary>missing methods, warning</primary></indexterm>
817 <indexterm><primary>methods, missing</primary></indexterm>
819 <para>This option is on by default, and warns you whenever
820 an instance declaration is missing one or more methods, and
821 the corresponding class declaration has no default
822 declaration for them.</para>
823 <para>The warning is suppressed if the method name
824 begins with an underscore. Here's an example where this is useful:
827 _simpleFn :: a -> String
828 complexFn :: a -> a -> String
829 complexFn x y = ... _simpleFn ...
831 The idea is that: (a) users of the class will only call <literal>complexFn</literal>;
832 never <literal>_simpleFn</literal>; and (b)
833 instance declarations can define either <literal>complexFn</literal> or <literal>_simpleFn</literal>.
839 <term><option>-fwarn-missing-signatures</option>:</term>
841 <indexterm><primary><option>-fwarn-missing-signatures</option></primary></indexterm>
842 <indexterm><primary>type signatures, missing</primary></indexterm>
844 <para>If you would like GHC to check that every top-level
845 function/value has a type signature, use the
846 <option>-fwarn-missing-signatures</option> option. This
847 option is off by default.</para>
852 <term><option>-fwarn-name-shadowing</option>:</term>
854 <indexterm><primary><option>-fwarn-name-shadowing</option></primary></indexterm>
855 <indexterm><primary>shadowing, warning</primary></indexterm>
857 <para>This option causes a warning to be emitted whenever an
858 inner-scope value has the same name as an outer-scope value,
859 i.e. the inner value shadows the outer one. This can catch
860 typographical errors that turn into hard-to-find bugs, e.g.,
861 in the inadvertent cyclic definition <literal>let x = ... x
862 ... in</literal>.</para>
864 <para>Consequently, this option does
865 <emphasis>will</emphasis> complain about cyclic recursive
871 <term><option>-fwarn-overlapping-patterns</option>:</term>
872 <indexterm><primary><option>-fwarn-overlapping-patterns</option></primary></indexterm>
873 <indexterm><primary>overlapping patterns, warning</primary></indexterm>
874 <indexterm><primary>patterns, overlapping</primary></indexterm>
876 <para>By default, the compiler will warn you if a set of
877 patterns are overlapping, i.e.,</para>
880 f :: String -> Int
886 <para>where the last pattern match in <Function>f</Function>
887 won't ever be reached, as the second pattern overlaps
888 it. More often than not, redundant patterns is a programmer
889 mistake/error, so this option is enabled by default.</para>
894 <term><option>-fwarn-simple-patterns</option>:</term>
896 <indexterm><primary><option>-fwarn-simple-patterns</option></primary>
898 <para>Causes the compiler to warn about lambda-bound
899 patterns that can fail, eg. <literal>\(x:xs)->...</literal>.
900 Normally, these aren't treated as incomplete patterns by
901 <option>-fwarn-incomplete-patterns</option>.</para>
902 <para>``Lambda-bound patterns'' includes all places where there is a single pattern,
903 including list comprehensions and do-notation. In these cases, a pattern-match
904 failure is quite legitimate, and triggers filtering (list comprehensions) or
905 the monad <literal>fail</literal> operation (monads). For example:
907 f :: [Maybe a] -> [a]
908 f xs = [y | Just y <- xs]
910 Switching on <option>-fwarn-simple-patterns</option> will elicit warnings about
911 these probably-innocent cases, which is why the flag is off by default. </para>
912 <para> The <literal>deriving( Read )</literal> mechanism produces monadic code with
913 pattern matches, so you will also get misleading warnings about the compiler-generated
914 code. (This is arguably a Bad Thing, but it's awkward to fix.)</para>
920 <term><option>-fwarn-type-defaults</option>:</term>
922 <indexterm><primary><option>-fwarn-type-defaults</option></primary></indexterm>
923 <indexterm><primary>defaulting mechanism, warning</primary></indexterm>
924 <para>Have the compiler warn/inform you where in your source
925 the Haskell defaulting mechanism for numeric types kicks
926 in. This is useful information when converting code from a
927 context that assumed one default into one with another,
928 e.g., the `default default' for Haskell 1.4 caused the
929 otherwise unconstrained value <Constant>1</Constant> to be
930 given the type <literal>Int</literal>, whereas Haskell 98
931 defaults it to <literal>Integer</literal>. This may lead to
932 differences in performance and behaviour, hence the
933 usefulness of being non-silent about this.</para>
935 <para>This warning is off by default.</para>
940 <term><option>-fwarn-unused-binds</option>:</term>
942 <indexterm><primary><option>-fwarn-unused-binds</option></primary></indexterm>
943 <indexterm><primary>unused binds, warning</primary></indexterm>
944 <indexterm><primary>binds, unused</primary></indexterm>
945 <para>Report any function definitions (and local bindings)
946 which are unused. For top-level functions, the warning is
947 only given if the binding is not exported.</para>
952 <term><option>-fwarn-unused-imports</option>:</term>
954 <indexterm><primary><option>-fwarn-unused-imports</option></primary></indexterm>
955 <indexterm><primary>unused imports, warning</primary></indexterm>
956 <indexterm><primary>imports, unused</primary></indexterm>
958 <para>Report any objects that are explicitly imported but
964 <term><option>-fwarn-unused-matches</option>:</term>
966 <indexterm><primary><option>-fwarn-unused-matches</option></primary></indexterm>
967 <indexterm><primary>unused matches, warning</primary></indexterm>
968 <indexterm><primary>matches, unused</primary></indexterm>
970 <para>Report all unused variables which arise from pattern
971 matches, including patterns consisting of a single variable.
972 For instance <literal>f x y = []</literal> would report
973 <VarName>x</VarName> and <VarName>y</VarName> as unused. The
974 warning is suppressed if the variable name begins with an underscore, thus:
984 <para>If you're feeling really paranoid, the
985 <option>-dcore-lint</option>
986 option<indexterm><primary><option>-dcore-lint</option></primary></indexterm>
987 is a good choice. It turns on heavyweight intra-pass
988 sanity-checking within GHC. (It checks GHC's sanity, not
995 <sect1 id="options-optimise">
996 <title>Optimisation (code improvement)</title>
998 <indexterm><primary>optimisation</primary></indexterm>
999 <indexterm><primary>improvement, code</primary></indexterm>
1001 <para>The <option>-O*</option> options specify convenient
1002 “packages” of optimisation flags; the
1003 <option>-f*</option> options described later on specify
1004 <emphasis>individual</emphasis> optimisations to be turned on/off;
1005 the <option>-m*</option> options specify
1006 <emphasis>machine-specific</emphasis> optimisations to be turned
1009 <sect2 id="optimise-pkgs">
1010 <title><option>-O*</option>: convenient “packages” of optimisation flags.</title>
1012 <para>There are <emphasis>many</emphasis> options that affect
1013 the quality of code produced by GHC. Most people only have a
1014 general goal, something like “Compile quickly” or
1015 “Make my program run like greased lightning.” The
1016 following “packages” of optimisations (or lack
1017 thereof) should suffice.</para>
1019 <para>Once you choose a <option>-O*</option>
1020 “package,” stick with it—don't chop and
1021 change. Modules' interfaces <emphasis>will</emphasis> change
1022 with a shift to a new <option>-O*</option> option, and you may
1023 have to recompile a large chunk of all importing modules before
1024 your program can again be run safely (see <XRef
1025 LinkEnd="recomp">).</para>
1030 <term>No <option>-O*</option>-type option specified:</term>
1031 <indexterm><primary>-O* not specified</primary></indexterm>
1033 <para>This is taken to mean: “Please compile
1034 quickly; I'm not over-bothered about compiled-code
1035 quality.” So, for example: <command>ghc -c
1036 Foo.hs</command></para>
1041 <term><option>-O0</option>:</term>
1042 <indexterm><primary><option>-O0</option></primary></indexterm>
1044 <para>Means “turn off all optimisation”,
1045 reverting to the same settings as if no
1046 <option>-O</option> options had been specified. Saying
1047 <option>-O0</option> can be useful if
1048 eg. <command>make</command> has inserted a
1049 <option>-O</option> on the command line already.</para>
1054 <term><option>-O</option> or <option>-O1</option>:</term>
1055 <indexterm><primary>-O option</primary></indexterm>
1056 <indexterm><primary>-O1 option</primary></indexterm>
1057 <indexterm><primary>optimise</primary><secondary>normally</secondary></indexterm>
1059 <para>Means: “Generate good-quality code without
1060 taking too long about it.” Thus, for example:
1061 <command>ghc -c -O Main.lhs</command></para>
1066 <term><option>-O2</option>:</term>
1067 <indexterm><primary>-O2 option</primary></indexterm>
1068 <indexterm><primary>optimise</primary><secondary>aggressively</secondary></indexterm>
1070 <para>Means: “Apply every non-dangerous
1071 optimisation, even if it means significantly longer
1072 compile times.”</para>
1074 <para>The avoided “dangerous” optimisations
1075 are those that can make runtime or space
1076 <emphasis>worse</emphasis> if you're unlucky. They are
1077 normally turned on or off individually.</para>
1079 <para>At the moment, <option>-O2</option> is
1080 <emphasis>unlikely</emphasis> to produce better code than
1081 <option>-O</option>.</para>
1086 <term><option>-Ofile <file></option>:</term>
1087 <indexterm><primary>-Ofile <file> option</primary></indexterm>
1088 <indexterm><primary>optimising, customised</primary></indexterm>
1090 <para>(NOTE: not supported yet in GHC 5.x. Please ask if
1091 you're interested in this.)</para>
1093 <para>For those who need <emphasis>absolute</emphasis>
1094 control over <emphasis>exactly</emphasis> what options are
1095 used (e.g., compiler writers, sometimes :-), a list of
1096 options can be put in a file and then slurped in with
1097 <option>-Ofile</option>.</para>
1099 <para>In that file, comments are of the
1100 <literal>#</literal>-to-end-of-line variety; blank
1101 lines and most whitespace is ignored.</para>
1103 <para>Please ask if you are baffled and would like an
1104 example of <option>-Ofile</option>!</para>
1109 <para>We don't use a <option>-O*</option> flag for day-to-day
1110 work. We use <option>-O</option> to get respectable speed;
1111 e.g., when we want to measure something. When we want to go for
1112 broke, we tend to use <option>-O -fvia-C</option> (and we go for
1113 lots of coffee breaks).</para>
1115 <para>The easiest way to see what <option>-O</option> (etc.)
1116 “really mean” is to run with <option>-v</option>,
1117 then stand back in amazement.</para>
1120 <sect2 id="options-f">
1121 <title><option>-f*</option>: platform-independent flags</title>
1123 <indexterm><primary>-f* options (GHC)</primary></indexterm>
1124 <indexterm><primary>-fno-* options (GHC)</primary></indexterm>
1126 <para>These flags turn on and off individual optimisations.
1127 They are normally set via the <option>-O</option> options
1128 described above, and as such, you shouldn't need to set any of
1129 them explicitly (indeed, doing so could lead to unexpected
1130 results). However, there are one or two that may be of
1135 <term><option>-fexcess-precision</option>:</term>
1137 <indexterm><primary><option>-fexcess-precision</option></primary></indexterm>
1138 <para>When this option is given, intermediate floating
1139 point values can have a <emphasis>greater</emphasis>
1140 precision/range than the final type. Generally this is a
1141 good thing, but some programs may rely on the exact
1143 <literal>Float</literal>/<literal>Double</literal> values
1144 and should not use this option for their compilation.</para>
1149 <term><option>-fignore-asserts</option>:</term>
1151 <indexterm><primary><option>-fignore-asserts</option></primary></indexterm>
1152 <para>Causes GHC to ignore uses of the function
1153 <literal>Exception.assert</literal> in source code (in
1154 other words, rewriting <literal>Exception.assert p
1155 e</literal> to <literal>e</literal> (see <xref
1156 linkend="sec-assertions">). This flag is turned on by
1157 <option>-O</option>.
1163 <term><option>-fno-strictness</option></term>
1164 <indexterm><primary><option>-fno-strictness</option></primary>
1167 <para>Turns off the strictness analyser; sometimes it eats
1168 too many cycles.</para>
1173 <term><option>-fno-cpr-analyse</option></term>
1174 <indexterm><primary><option>-fno-cpr-analyse</option></primary>
1177 <para>Turns off the CPR (constructed product result)
1178 analysis; it is somewhat experimental.</para>
1183 <term><option>-funbox-strict-fields</option>:</term>
1185 <indexterm><primary><option>-funbox-strict-fields</option></primary></indexterm>
1186 <indexterm><primary>strict constructor fields</primary></indexterm>
1187 <indexterm><primary>constructor fields, strict</primary></indexterm>
1189 <para>This option causes all constructor fields which are
1190 marked strict (i.e. “!”) to be unboxed or
1191 unpacked if possible. For example:</para>
1194 data T = T !Float !Float
1197 <para>will create a constructor <literal>T</literal>
1198 containing two unboxed floats if the
1199 <option>-funbox-strict-fields</option> flag is given.
1200 This may not always be an optimisation: if the
1201 <Function>T</Function> constructor is scrutinised and the
1202 floats passed to a non-strict function for example, they
1203 will have to be reboxed (this is done automatically by the
1206 <para>This option should only be used in conjunction with
1207 <option>-O</option>, in order to expose unfoldings to the
1208 compiler so the reboxing can be removed as often as
1209 possible. For example:</para>
1213 f (T f1 f2) = f1 + f2
1216 <para>The compiler will avoid reboxing
1217 <Function>f1</Function> and <Function>f2</Function> by
1218 inlining <Function>+</Function> on floats, but only when
1219 <option>-O</option> is on.</para>
1221 <para>Any single-constructor data is eligible for
1222 unpacking; for example</para>
1225 data T = T !(Int,Int)
1228 <para>will store the two <literal>Int</literal>s directly
1229 in the <Function>T</Function> constructor, by flattening
1230 the pair. Multi-level unpacking is also supported:</para>
1234 data S = S !Int !Int
1237 <para>will store two unboxed <literal>Int#</literal>s
1238 directly in the <Function>T</Function> constructor.</para>
1243 <term><option>-funfolding-update-in-place<n></option></term>
1244 <indexterm><primary><option>-funfolding-update-in-place</option></primary></indexterm>
1246 <para>Switches on an experimental "optimisation".
1247 Switching it on makes the compiler a little keener to
1248 inline a function that returns a constructor, if the
1249 context is that of a thunk.
1253 If we inlined plusInt we might get an opportunity to use
1254 update-in-place for the thunk 'x'.</para>
1259 <term><option>-funfolding-creation-threshold<n></option>:</term>
1261 <indexterm><primary><option>-funfolding-creation-threshold</option></primary></indexterm>
1262 <indexterm><primary>inlining, controlling</primary></indexterm>
1263 <indexterm><primary>unfolding, controlling</primary></indexterm>
1265 <para>(Default: 45) Governs the maximum size that GHC will
1266 allow a function unfolding to be. (An unfolding has a
1267 “size” that reflects the cost in terms of
1268 “code bloat” of expanding that unfolding at
1269 at a call site. A bigger function would be assigned a
1270 bigger cost.) </para>
1272 <para> Consequences: (a) nothing larger than this will be
1273 inlined (unless it has an INLINE pragma); (b) nothing
1274 larger than this will be spewed into an interface
1278 <para> Increasing this figure is more likely to result in longer
1279 compile times than faster code. The next option is more
1285 <term><option>-funfolding-use-threshold<n></option>:</term>
1287 <indexterm><primary><option>-funfolding-use-threshold</option></primary></indexterm>
1288 <indexterm><primary>inlining, controlling</primary></indexterm>
1289 <indexterm><primary>unfolding, controlling</primary></indexterm>
1291 <para>(Default: 8) This is the magic cut-off figure for
1292 unfolding: below this size, a function definition will be
1293 unfolded at the call-site, any bigger and it won't. The
1294 size computed for a function depends on two things: the
1295 actual size of the expression minus any discounts that
1296 apply (see <option>-funfolding-con-discount</option>).</para>
1307 <sect1 id="sec-using-concurrent">
1308 <title>Using Concurrent Haskell</title>
1310 <indexterm><primary>Concurrent Haskell—use</primary></indexterm>
1313 GHC supports Concurrent Haskell by default, without requiring a
1314 special option or libraries compiled in a certain way. To get access
1315 to the support libraries for Concurrent Haskell, just import
1316 <literal>Control.Concurrent</literal> (details are in the accompanying
1317 library documentation).</para>
1320 RTS options are provided for modifying the behaviour of the threaded
1321 runtime system. See <XRef LinkEnd="parallel-rts-opts">.
1325 Concurrent Haskell is described in more detail in the documentation
1326 for the <literal>Control.Concurrent</literal> module.
1331 <Sect1 id="sec-using-parallel">
1332 <title>Using Parallel Haskell</title>
1335 <indexterm><primary>Parallel Haskell—use</primary></indexterm>
1339 [You won't be able to execute parallel Haskell programs unless PVM3
1340 (Parallel Virtual Machine, version 3) is installed at your site.]
1344 To compile a Haskell program for parallel execution under PVM, use the
1345 <Option>-parallel</Option> option,<IndexTerm><Primary>-parallel
1346 option</Primary></IndexTerm> both when compiling <Emphasis>and
1347 linking</Emphasis>. You will probably want to <Literal>import
1348 Parallel</Literal> into your Haskell modules.
1352 To run your parallel program, once PVM is going, just invoke it
1353 “as normal”. The main extra RTS option is
1354 <Option>-qp<n></Option>, to say how many PVM
1355 “processors” your program to run on. (For more details of
1356 all relevant RTS options, please see <XRef
1357 LinkEnd="parallel-rts-opts">.)
1361 In truth, running Parallel Haskell programs and getting information
1362 out of them (e.g., parallelism profiles) is a battle with the vagaries of
1363 PVM, detailed in the following sections.
1366 <Sect2 id="pvm-dummies">
1367 <Title>Dummy's guide to using PVM</Title>
1370 <indexterm><primary>PVM, how to use</primary></indexterm>
1371 <indexterm><primary>Parallel Haskell—PVM use</primary></indexterm>
1372 Before you can run a parallel program under PVM, you must set the
1373 required environment variables (PVM's idea, not ours); something like,
1374 probably in your <filename>.cshrc</filename> or equivalent:
1377 setenv PVM_ROOT /wherever/you/put/it
1378 setenv PVM_ARCH `$PVM_ROOT/lib/pvmgetarch`
1379 setenv PVM_DPATH $PVM_ROOT/lib/pvmd
1385 Creating and/or controlling your “parallel machine” is a purely-PVM
1386 business; nothing specific to Parallel Haskell. The following paragraphs
1387 describe how to configure your parallel machine interactively.
1391 If you use parallel Haskell regularly on the same machine configuration it
1392 is a good idea to maintain a file with all machine names and to make the
1393 environment variable PVM_HOST_FILE point to this file. Then you can avoid
1394 the interactive operations described below by just saying
1402 You use the <Command>pvm</Command><IndexTerm><Primary>pvm command</Primary></IndexTerm> command to start PVM on your
1403 machine. You can then do various things to control/monitor your
1404 “parallel machine;” the most useful being:
1410 <ColSpec Align="Left">
1414 <entry><KeyCombo><KeyCap>Control</KeyCap><KeyCap>D</KeyCap></KeyCombo></entry>
1415 <entry>exit <command>pvm</command>, leaving it running</entry>
1419 <entry><command>halt</command></entry>
1420 <entry>kill off this “parallel machine” & exit</entry>
1424 <entry><command>add <host></command></entry>
1425 <entry>add <command><host></command> as a processor</entry>
1429 <entry><command>delete <host></command></entry>
1430 <entry>delete <command><host></command></entry>
1434 <entry><command>reset</command></entry>
1435 <entry>kill what's going, but leave PVM up</entry>
1439 <entry><command>conf</command></entry>
1440 <entry>list the current configuration</entry>
1444 <entry><command>ps</command></entry>
1445 <entry>report processes' status</entry>
1449 <entry><command>pstat <pid></command></entry>
1450 <entry>status of a particular process</entry>
1459 The PVM documentation can tell you much, much more about <command>pvm</command>!
1464 <Sect2 id="par-profiles">
1465 <Title>Parallelism profiles</Title>
1468 <indexterm><primary>parallelism profiles</primary></indexterm>
1469 <indexterm><primary>profiles, parallelism</primary></indexterm>
1470 <indexterm><primary>visualisation tools</primary></indexterm>
1474 With Parallel Haskell programs, we usually don't care about the
1475 results—only with “how parallel” it was! We want pretty pictures.
1479 Parallelism profiles (à la <Command>hbcpp</Command>) can be generated with the
1480 <Option>-qP</Option><IndexTerm><Primary>-qP RTS option (concurrent, parallel)</Primary></IndexTerm> RTS option. The
1481 per-processor profiling info is dumped into files named
1482 <Filename><full-path><program>.gr</Filename>. These are then munged into a PostScript picture,
1483 which you can then display. For example, to run your program
1484 <Filename>a.out</Filename> on 8 processors, then view the parallelism profile, do:
1490 <prompt>$</prompt> ./a.out +RTS -qP -qp8
1491 <prompt>$</prompt> grs2gr *.???.gr > temp.gr # combine the 8 .gr files into one
1492 <prompt>$</prompt> gr2ps -O temp.gr # cvt to .ps; output in temp.ps
1493 <prompt>$</prompt> ghostview -seascape temp.ps # look at it!
1499 The scripts for processing the parallelism profiles are distributed
1500 in <filename>ghc/utils/parallel/</filename>.
1506 <Title>Other useful info about running parallel programs</Title>
1509 The “garbage-collection statistics” RTS options can be useful for
1510 seeing what parallel programs are doing. If you do either
1511 <Option>+RTS -Sstderr</Option><IndexTerm><Primary>-Sstderr RTS option</Primary></IndexTerm> or <Option>+RTS -sstderr</Option>, then
1512 you'll get mutator, garbage-collection, etc., times on standard
1513 error. The standard error of all PE's other than the `main thread'
1514 appears in <filename>/tmp/pvml.nnn</filename>, courtesy of PVM.
1518 Whether doing <option>+RTS -Sstderr</option> or not, a handy way to watch
1519 what's happening overall is: <command>tail -f /tmp/pvml.nnn</command>.
1524 <Sect2 id="parallel-rts-opts">
1525 <title>RTS options for Concurrent/Parallel Haskell
1529 <indexterm><primary>RTS options, concurrent</primary></indexterm>
1530 <indexterm><primary>RTS options, parallel</primary></indexterm>
1531 <indexterm><primary>Concurrent Haskell—RTS options</primary></indexterm>
1532 <indexterm><primary>Parallel Haskell—RTS options</primary></indexterm>
1536 Besides the usual runtime system (RTS) options
1537 (<XRef LinkEnd="runtime-control">), there are a few options particularly
1538 for concurrent/parallel execution.
1545 <Term><Option>-qp<N></Option>:</Term>
1548 <IndexTerm><Primary>-qp<N> RTS option</Primary></IndexTerm>
1549 (PARALLEL ONLY) Use <Literal><N></Literal> PVM processors to run this program;
1555 <term><option>-C[<us>]</option>:</term>
1558 <indexterm><primary>-C<us> RTS option</primary></indexterm> Sets
1559 the context switch interval to <literal><s></literal> seconds.
1560 A context switch will occur at the next heap block allocation after
1561 the timer expires (a heap block allocation occurs every 4k of
1562 allocation). With <option>-C0</option> or <option>-C</option>,
1563 context switches will occur as often as possible (at every heap block
1564 allocation). By default, context switches occur every 20ms
1565 milliseconds. Note that GHC's internal timer ticks every 20ms, and
1566 the context switch timer is always a multiple of this timer, so 20ms
1567 is the maximum granularity available for timed context switches.
1572 <term><option>-q[v]</option>:</term>
1575 <indexterm><primary>-q RTS option</primary></indexterm>
1576 (PARALLEL ONLY) Produce a quasi-parallel profile of thread activity,
1577 in the file <FIlename><program>.qp</FIlename>. In the style of <command>hbcpp</command>, this profile
1578 records the movement of threads between the green (runnable) and red
1579 (blocked) queues. If you specify the verbose suboption (<option>-qv</option>), the
1580 green queue is split into green (for the currently running thread
1581 only) and amber (for other runnable threads). We do not recommend
1582 that you use the verbose suboption if you are planning to use the
1583 <Command>hbcpp</Command> profiling tools or if you are context switching at every heap
1584 check (with <Option>-C</Option>).
1590 <Term><Option>-qt<num></Option>:</Term>
1593 <IndexTerm><Primary>-qt<num> RTS option</Primary></IndexTerm>
1594 (PARALLEL ONLY) Limit the thread pool size, i.e. the number of concurrent
1595 threads per processor to <Literal><num></Literal>. The default is
1596 32. Each thread requires slightly over 1K <Emphasis>words</Emphasis> in
1597 the heap for thread state and stack objects. (For 32-bit machines, this
1598 translates to 4K bytes, and for 64-bit machines, 8K bytes.)
1604 <Term><Option>-d</Option>:</Term>
1607 <IndexTerm><Primary>-d RTS option (parallel)</Primary></IndexTerm>
1608 (PARALLEL ONLY) Turn on debugging. It pops up one xterm (or GDB, or
1609 something…) per PVM processor. We use the standard <Command>debugger</Command>
1610 script that comes with PVM3, but we sometimes meddle with the
1611 <Command>debugger2</Command> script. We include ours in the GHC distribution,
1612 in <Filename>ghc/utils/pvm/</Filename>.
1618 <Term><Option>-qe<num></Option>:</Term>
1621 <IndexTerm><Primary>-qe<num> RTS option
1622 (parallel)</Primary></IndexTerm> (PARALLEL ONLY) Limit the spark pool size
1623 i.e. the number of pending sparks per processor to
1624 <Literal><num></Literal>. The default is 100. A larger number may be
1625 appropriate if your program generates large amounts of parallelism
1631 <Term><Option>-qQ<num></Option>:</Term>
1634 <IndexTerm><Primary>-qQ<num> RTS option (parallel)</Primary></IndexTerm>
1635 (PARALLEL ONLY) Set the size of packets transmitted between processors
1636 to <Literal><num></Literal>. The default is 1024 words. A larger number may be
1637 appropriate if your machine has a high communication cost relative to
1643 <Term><Option>-qh<num></Option>:</Term>
1646 <IndexTerm><Primary>-qh<num> RTS option (parallel)</Primary></IndexTerm>
1647 (PARALLEL ONLY) Select a packing scheme. Set the number of non-root thunks to pack in one packet to
1648 <num>-1 (0 means infinity). By default GUM uses full-subgraph
1649 packing, i.e. the entire subgraph with the requested closure as root is
1650 transmitted (provided it fits into one packet). Choosing a smaller value
1651 reduces the amount of pre-fetching of work done in GUM. This can be
1652 advantageous for improving data locality but it can also worsen the balance
1653 of the load in the system.
1658 <Term><Option>-qg<num></Option>:</Term>
1661 <IndexTerm><Primary>-qg<num> RTS option
1662 (parallel)</Primary></IndexTerm> (PARALLEL ONLY) Select a globalisation
1663 scheme. This option affects the
1664 generation of global addresses when transferring data. Global addresses are
1665 globally unique identifiers required to maintain sharing in the distributed
1666 graph structure. Currently this is a binary option. With <num>=0 full globalisation is used
1667 (default). This means a global address is generated for every closure that
1668 is transmitted. With <num>=1 a thunk-only globalisation scheme is
1669 used, which generated global address only for thunks. The latter case may
1670 lose sharing of data but has a reduced overhead in packing graph structures
1671 and maintaining internal tables of global addresses.
1682 <sect1 id="options-platform">
1683 <title>Platform-specific Flags</title>
1685 <indexterm><primary>-m* options</primary></indexterm>
1686 <indexterm><primary>platform-specific options</primary></indexterm>
1687 <indexterm><primary>machine-specific options</primary></indexterm>
1689 <para>Some flags only make sense for particular target
1695 <term><option>-mv8</option>:</term>
1697 <para>(SPARC machines)<indexterm><primary>-mv8 option (SPARC
1698 only)</primary></indexterm> Means to pass the like-named
1699 option to GCC; it says to use the Version 8 SPARC
1700 instructions, notably integer multiply and divide. The
1701 similiar <option>-m*</option> GCC options for SPARC also
1702 work, actually.</para>
1707 <term><option>-monly-[32]-regs</option>:</term>
1709 <para>(iX86 machines)<indexterm><primary>-monly-N-regs
1710 option (iX86 only)</primary></indexterm> GHC tries to
1711 “steal” four registers from GCC, for performance
1712 reasons; it almost always works. However, when GCC is
1713 compiling some modules with four stolen registers, it will
1714 crash, probably saying:
1717 Foo.hc:533: fixed or forbidden register was spilled.
1718 This may be due to a compiler bug or to impossible asm
1719 statements or clauses.
1722 Just give some registers back with
1723 <option>-monly-N-regs</option>. Try `3' first, then `2'.
1724 If `2' doesn't work, please report the bug to us.</para>
1733 <sect1 id="ext-core">
1734 <title>Generating and compiling External Core Files</title>
1736 <indexterm><primary>intermediate code generation</primary></indexterm>
1738 <para>GHC can dump its optimized intermediate code (said to be in “Core” format)
1739 to a file as a side-effect of compilation. Core files, which are given the suffix
1740 <filename>.hcr</filename>, can be read and processed by non-GHC back-end
1741 tools. The Core format is formally described in <ulink url="http://www.haskell.org/ghc/docs/papers/core.ps.gz"
1742 <citetitle>An External Representation for the GHC Core Language</citetitle></ulink>,
1743 and sample tools (in Haskell)
1744 for manipulating Core files are available in the GHC source distribution
1745 directory <literal>/fptools/ghc/utils/ext-core</literal>.
1746 Note that the format of <literal>.hcr</literal>
1747 files is <emphasis>different</emphasis> (though similar) to the Core output format generated
1748 for debugging purposes (<xref linkend="options-debugging">).</para>
1750 <para>The Core format natively supports notes which you can add to
1751 your source code using the <literal>CORE</literal> pragma (see <xref
1752 linkend="pragmas">).</para>
1757 <term><option>-fext-core</option></term>
1759 <primary><option>-fext-core</option></primary>
1762 <para>Generate <literal>.hcr</literal> files.</para>
1768 <para>GHC can also read in External Core files as source; just give the <literal>.hcr</literal> file on
1769 the command line, instead of the <literal>.hs</literal> or <literal>.lhs</literal> Haskell source.
1770 A current infelicity is that you need to give teh <literal>-fglasgow-exts</literal> flag too, because
1771 ordinary Haskell 98, when translated to External Core, uses things like rank-2 types.</para>
1780 ;;; Local Variables: ***
1782 ;;; sgml-parent-document: ("users_guide.sgml" "book" "chapter") ***