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, options which relate to filenames are
109 static, and the rest are dynamic. The flag reference tables (<xref
110 linkend="flag-reference"/>) lists the status of each flag.</para>
113 <sect1 id="file-suffixes">
114 <title>Meaningful file suffixes</title>
116 <indexterm><primary>suffixes, file</primary></indexterm>
117 <indexterm><primary>file suffixes for GHC</primary></indexterm>
119 <para>File names with “meaningful” suffixes (e.g.,
120 <filename>.lhs</filename> or <filename>.o</filename>) cause the
121 “right thing” to happen to those files.</para>
126 <term><filename>.lhs</filename></term>
127 <indexterm><primary><literal>lhs</literal> suffix</primary></indexterm>
129 <para>A “literate Haskell” module.</para>
134 <term><filename>.hs</filename></term>
136 <para>A not-so-literate Haskell module.</para>
141 <term><filename>.hi</filename></term>
143 <para>A Haskell interface file, probably
144 compiler-generated.</para>
149 <term><filename>.hc</filename></term>
151 <para>Intermediate C file produced by the Haskell
157 <term><filename>.c</filename></term>
159 <para>A C file not produced by the Haskell
165 <term><filename>.s</filename></term>
167 <para>An assembly-language source file, usually produced by
173 <term><filename>.o</filename></term>
175 <para>An object file, produced by an assembler.</para>
180 <para>Files with other suffixes (or without suffixes) are passed
181 straight to the linker.</para>
186 <title>Modes of operation</title>
188 <para>GHC's behaviour is firstly controlled by a mode flag. Only
189 one of these flags may be given, but it does not necessarily need
190 to be the first option on the command-line. The available modes
195 <term><cmdsynopsis><command>ghc</command>
196 <arg choice='plain'>––interactive</arg>
197 </cmdsynopsis></term>
198 <indexterm><primary>interactive mode</primary>
200 <indexterm><primary>ghci</primary>
203 <para>Interactive mode, which is also available as
204 <command>ghci</command>. Interactive mode is described in
205 more detail in <xref linkend="ghci"/>.</para>
210 <term><cmdsynopsis><command>ghc</command>
211 <arg choice='plain'>––make</arg>
212 </cmdsynopsis></term>
213 <indexterm><primary>make mode</primary>
215 <indexterm><primary><option>––make</option></primary>
218 <para>In this mode, GHC will build a multi-module Haskell
219 program automatically, figuring out dependencies for itself.
220 If you have a straightforward Haskell program, this is
221 likely to be much easier, and faster, than using
222 <command>make</command>. Make mode is described in <xref
223 linkend="make-mode"/>.</para>
228 <term><cmdsynopsis><command>ghc</command>
229 <arg choice='plain'>–e</arg> <arg choice='plain'><replaceable>expr</replaceable></arg>
230 </cmdsynopsis></term>
231 <indexterm><primary>eval mode</primary>
234 <para>Expression-evaluation mode. This is very similar to
235 interactive mode, except that there is a single expression
236 to evaluate (<replaceable>expr</replaceable>) which is given
237 on the command line. See <xref linkend="eval-mode"/> for
244 <command>ghc</command>
251 </cmdsynopsis></term>
252 <indexterm><primary><option>-E</option></primary></indexterm>
253 <indexterm><primary><option>-C</option></primary></indexterm>
254 <indexterm><primary><option>-S</option></primary></indexterm>
255 <indexterm><primary><option>-c</option></primary></indexterm>
257 <para>This is the traditional batch-compiler mode, in which
258 GHC can compile source files one at a time, or link objects
259 together into an executable. This mode also applies if
260 there is no other mode flag specified on the command line,
261 in which case it means that the specified files should be
262 compiled and then linked to form a program. See <xref
263 linkend="options-order"/>.</para>
268 <term><cmdsynopsis><command>ghc</command>
269 <arg choice='plain'>–M</arg></cmdsynopsis></term>
270 <indexterm><primary>dependency-generation mode</primary>
273 <para>Dependency-generation mode. In this mode, GHC can be
274 used to generate dependency information suitable for use in
275 a <literal>Makefile</literal>. See <xref
276 linkend="sec-makefile-dependencies"/>.</para>
281 <term><cmdsynopsis><command>ghc</command>
282 <arg choice='plain'>––mk-dll</arg></cmdsynopsis></term>
283 <indexterm><primary>dependency-generation mode</primary>
286 <para>DLL-creation mode (Windows only). See <xref
287 linkend="win32-dlls-create"/>.</para>
292 <sect2 id="make-mode">
293 <title>Using <command>ghc</command> <option>––make</option></title>
295 <indexterm><primary><option>––make</option></primary>
297 <indexterm><primary>separate compilation</primary>
300 <para>When given the <option>––make</option> option,
301 GHC will build a multi-module Haskell program by following
302 dependencies from a single root module (usually
303 <literal>Main</literal>). For example, if your
304 <literal>Main</literal> module is in a file called
305 <filename>Main.hs</filename>, you could compile and link the
306 program like this:</para>
309 ghc ––make Main.hs
312 <para>The command line may contain any number of source file
313 names or module names; GHC will figure out all the modules in
314 the program by following the imports from these initial modules.
315 It will then attempt to compile each module which is out of
316 date, and finally if there is a <literal>Main</literal> module,
317 the program will also be linked into an executable.</para>
319 <para>The main advantages to using <literal>ghc
320 ––make</literal> over traditional
321 <literal>Makefile</literal>s are:</para>
325 <para>GHC doesn't have to be restarted for each compilation,
326 which means it can cache information between compilations.
327 Compiling a muli-module program with <literal>ghc
328 ––make</literal> can be up to twice as fast as
329 running <literal>ghc</literal> individually on each source
333 <para>You don't have to write a
334 <literal>Makefile</literal>.</para>
336 <indexterm><primary><literal>Makefile</literal>s</primary><secondary>avoiding</secondary>
339 <para>GHC re-calculates the dependencies each time it is
340 invoked, so the dependencies never get out of sync with the
345 <para>Any of the command-line options described in the rest of
346 this chapter can be used with
347 <option>––make</option>, but note that any options
348 you give on the command line will apply to all the source files
349 compiled, so if you want any options to apply to a single source
350 file only, you'll need to use an <literal>OPTIONS</literal>
351 pragma (see <xref linkend="source-file-options"/>).</para>
353 <para>If the program needs to be linked with additional objects
354 (say, some auxilliary C code), then the object files can be
355 given on the command line and GHC will include them when linking
356 the executable.</para>
358 <para>Note that GHC can only follow dependencies if it has the
359 source file available, so if your program includes a module for
360 which there is no source file, even if you have an object and an
361 interface file for the module, then GHC will complain. The
362 exception to this rule is for package modules, which may or may
363 not have source files.</para>
365 <para>The source files for the program don't all need to be in
366 the same directory; the <option>-i</option> option can be used
367 to add directories to the search path (see <xref
368 linkend="search-path"/>).</para>
371 <sect2 id="eval-mode">
372 <title>Expression evaluation mode</title>
374 <para>This mode is very similar to interactive mode, except that
375 there is a single expression to evaluate which is specified on
376 the command line as an argument to the <option>-e</option>
380 ghc -e <replaceable>expr</replaceable>
383 <para>Haskell source files may be named on the command line, and
384 they will be loaded exactly as in interactive mode. The
385 expression is evaluated in the context of the loaded
388 <para>For example, to load and run a Haskell program containing
389 a module <literal>Main</literal>, we might say</para>
392 ghc -e Main.main Main.hs
395 <para>or we can just use this mode to evaluate expressions in
396 the context of the <literal>Prelude</literal>:</para>
399 $ ghc -e "interact (unlines.map reverse.lines)"
405 <sect2 id="options-order">
406 <title>Batch compiler mode</title>
408 <para>In <emphasis>batch mode</emphasis>, GHC will compile one or more source files
409 given on the command line.</para>
411 <para>The first phase to run is determined by each input-file
412 suffix, and the last phase is determined by a flag. If no
413 relevant flag is present, then go all the way through linking.
414 This table summarises:</para>
418 <colspec align="left"/>
419 <colspec align="left"/>
420 <colspec align="left"/>
421 <colspec align="left"/>
425 <entry>Phase of the compilation system</entry>
426 <entry>Suffix saying “start here”</entry>
427 <entry>Flag saying “stop after”</entry>
428 <entry>(suffix of) output file</entry>
433 <entry>literate pre-processor</entry>
434 <entry><literal>.lhs</literal></entry>
436 <entry><literal>.hs</literal></entry>
440 <entry>C pre-processor (opt.) </entry>
441 <entry><literal>.hs</literal> (with
442 <option>-cpp</option>)</entry>
443 <entry><option>-E</option></entry>
444 <entry><literal>.hspp</literal></entry>
448 <entry>Haskell compiler</entry>
449 <entry><literal>.hs</literal></entry>
450 <entry><option>-C</option>, <option>-S</option></entry>
451 <entry><literal>.hc</literal>, <literal>.s</literal></entry>
455 <entry>C compiler (opt.)</entry>
456 <entry><literal>.hc</literal> or <literal>.c</literal></entry>
457 <entry><option>-S</option></entry>
458 <entry><literal>.s</literal></entry>
462 <entry>assembler</entry>
463 <entry><literal>.s</literal></entry>
464 <entry><option>-c</option></entry>
465 <entry><literal>.o</literal></entry>
469 <entry>linker</entry>
470 <entry><replaceable>other</replaceable></entry>
472 <entry><filename>a.out</filename></entry>
478 <indexterm><primary><option>-C</option></primary></indexterm>
479 <indexterm><primary><option>-E</option></primary></indexterm>
480 <indexterm><primary><option>-S</option></primary></indexterm>
481 <indexterm><primary><option>-c</option></primary></indexterm>
483 <para>Thus, a common invocation would be: </para>
486 ghc -c Foo.hs</screen>
488 <para>to compile the Haskell source file
489 <filename>Foo.hs</filename> to an object file
490 <filename>Foo.o</filename>.</para>
492 <para>Note: What the Haskell compiler proper produces depends on
493 whether a native-code generator<indexterm><primary>native-code
494 generator</primary></indexterm> is used (producing assembly
495 language) or not (producing C). See <xref
496 linkend="options-codegen"/> for more details.</para>
498 <para>Note: C pre-processing is optional, the
499 <option>-cpp</option><indexterm><primary><option>-cpp</option></primary></indexterm>
500 flag turns it on. See <xref linkend="c-pre-processor"/> for more
503 <para>Note: The option <option>-E</option><indexterm><primary>-E
504 option</primary></indexterm> runs just the pre-processing passes
505 of the compiler, dumping the result in a file. Note that this
506 differs from the previous behaviour of dumping the file to
507 standard output.</para>
511 <sect1 id="options-help">
512 <title>Help and verbosity options</title>
514 <indexterm><primary>help options</primary></indexterm>
515 <indexterm><primary>verbosity options</primary></indexterm>
519 <term><option>––help</option></term>
520 <term><option>-?</option></term>
521 <indexterm><primary><option>-?</option></primary></indexterm>
522 <indexterm><primary><option>––help</option></primary></indexterm>
524 <para>Cause GHC to spew a long usage message to standard
525 output and then exit.</para>
530 <term><option>-v</option></term>
531 <indexterm><primary><option>-v</option></primary></indexterm>
533 <para>The <option>-v</option> option makes GHC
534 <emphasis>verbose</emphasis>: it reports its version number
535 and shows (on stderr) exactly how it invokes each phase of
536 the compilation system. Moreover, it passes the
537 <option>-v</option> flag to most phases; each reports its
538 version number (and possibly some other information).</para>
540 <para>Please, oh please, use the <option>-v</option> option
541 when reporting bugs! Knowing that you ran the right bits in
542 the right order is always the first thing we want to
548 <term><option>-v</option><replaceable>n</replaceable></term>
549 <indexterm><primary><option>-v</option></primary></indexterm>
551 <para>To provide more control over the compiler's verbosity,
552 the <option>-v</option> flag takes an optional numeric
553 argument. Specifying <option>-v</option> on its own is
554 equivalent to <option>-v3</option>, and the other levels
555 have the following meanings:</para>
559 <term><option>-v0</option></term>
561 <para>Disable all non-essential messages (this is the
567 <term><option>-v1</option></term>
569 <para>Minimal verbosity: print one line per
570 compilation (this is the default when
571 <option>––make</option> or
572 <option>––interactive</option> is on).</para>
577 <term><option>-v2</option></term>
579 <para>Print the name of each compilation phase as it
580 is executed. (equivalent to
581 <option>-dshow-passes</option>).</para>
586 <term><option>-v3</option></term>
588 <para>The same as <option>-v2</option>, except that in
589 addition the full command line (if appropriate) for
590 each compilation phase is also printed.</para>
595 <term><option>-v4</option></term>
597 <para>The same as <option>-v3</option> except that the
598 intermediate program representation after each
599 compilation phase is also printed (excluding
600 preprocessed and C/assembly files).</para>
608 <term><option>-V</option></term>
609 <term><option>––version</option></term>
610 <indexterm><primary><option>-V</option></primary></indexterm>
611 <indexterm><primary><option>––version</option></primary></indexterm>
613 <para>Print a one-line string including GHC's version number.</para>
618 <term><option>––numeric-version</option></term>
619 <indexterm><primary><option>––numeric-version</option></primary></indexterm>
621 <para>Print GHC's numeric version number only.</para>
626 <term><option>––print-libdir</option></term>
627 <indexterm><primary><option>––print-libdir</option></primary></indexterm>
629 <para>Print the path to GHC's library directory. This is
630 the top of the directory tree containing GHC's libraries,
631 interfaces, and include files (usually something like
632 <literal>/usr/local/lib/ghc-5.04</literal> on Unix). This
634 <literal>$libdir</literal><indexterm><primary><literal>libdir</literal></primary>
635 </indexterm>in the package configuration file (see <xref
636 linkend="packages"/>).</para>
645 <sect1 id="options-sanity">
646 <title>Warnings and sanity-checking</title>
648 <indexterm><primary>sanity-checking options</primary></indexterm>
649 <indexterm><primary>warnings</primary></indexterm>
652 <para>GHC has a number of options that select which types of
653 non-fatal error messages, otherwise known as warnings, can be
654 generated during compilation. By default, you get a standard set
655 of warnings which are generally likely to indicate bugs in your
657 <option>-fwarn-overlapping-patterns</option>,
658 <option>-fwarn-deprecations</option>,
659 <option>-fwarn-duplicate-exports</option>,
660 <option>-fwarn-missing-fields</option>, and
661 <option>-fwarn-missing-methods</option>. The following flags are
662 simple ways to select standard “packages” of warnings:
668 <term><option>-W</option>:</term>
670 <indexterm><primary>-W option</primary></indexterm>
671 <para>Provides the standard warnings plus
672 <option>-fwarn-incomplete-patterns</option>,
673 <option>-fwarn-unused-matches</option>,
674 <option>-fwarn-unused-imports</option>,
675 <option>-fwarn-misc</option>, and
676 <option>-fwarn-unused-binds</option>.</para>
681 <term><option>-w</option>:</term>
683 <indexterm><primary><option>-w</option></primary></indexterm>
684 <para>Turns off all warnings, including the standard ones.</para>
689 <term><option>-Wall</option>:</term>
691 <indexterm><primary><option>-Wall</option></primary></indexterm>
692 <para>Turns on all warning options.</para>
697 <term><option>-Werror</option>:</term>
699 <indexterm><primary><option>-Werror</option></primary></indexterm>
700 <para>Makes any warning into a fatal error. Useful so that you don't
701 miss warnings when doing batch compilation. </para>
707 <para>The full set of warning options is described below. To turn
708 off any warning, simply give the corresponding
709 <option>-fno-warn-...</option> option on the command line.</para>
714 <term><option>-fwarn-deprecations</option>:</term>
716 <indexterm><primary><option>-fwarn-deprecations</option></primary>
718 <indexterm><primary>deprecations</primary></indexterm>
719 <para>Causes a warning to be emitted when a deprecated
720 function or type is used. Entities can be marked as
721 deprecated using a pragma, see <xref
722 linkend="deprecated-pragma"/>.</para>
727 <term><option>-fwarn-duplicate-exports</option>:</term>
729 <indexterm><primary><option>-fwarn-duplicate-exports</option></primary></indexterm>
730 <indexterm><primary>duplicate exports, warning</primary></indexterm>
731 <indexterm><primary>export lists, duplicates</primary></indexterm>
733 <para>Have the compiler warn about duplicate entries in
734 export lists. This is useful information if you maintain
735 large export lists, and want to avoid the continued export
736 of a definition after you've deleted (one) mention of it in
737 the export list.</para>
739 <para>This option is on by default.</para>
744 <term><option>-fwarn-hi-shadowing</option>:</term>
746 <indexterm><primary><option>-fwarn-hi-shadowing</option></primary></indexterm>
747 <indexterm><primary>shadowing</primary>
748 <secondary>interface files</secondary></indexterm>
750 <para>Causes the compiler to emit a warning when a module or
751 interface file in the current directory is shadowing one
752 with the same module name in a library or other
758 <term><option>-fwarn-incomplete-patterns</option>:</term>
760 <indexterm><primary><option>-fwarn-incomplete-patterns</option></primary></indexterm>
761 <indexterm><primary>incomplete patterns, warning</primary></indexterm>
762 <indexterm><primary>patterns, incomplete</primary></indexterm>
764 <para>Similarly for incomplete patterns, the function
765 <function>g</function> below will fail when applied to
766 non-empty lists, so the compiler will emit a warning about
767 this when <option>-fwarn-incomplete-patterns</option> is
774 <para>This option isn't enabled be default because it can be
775 a bit noisy, and it doesn't always indicate a bug in the
776 program. However, it's generally considered good practice
777 to cover all the cases in your functions.</para>
782 <term><option>-fwarn-misc</option>:</term>
783 <indexterm><primary><option>-fwarn-misc</option></primary></indexterm>
785 <para>Turns on warnings for various harmless but untidy
786 things. This currently includes: importing a type with
787 <literal>(..)</literal> when the export is abstract, and
788 listing duplicate class assertions in a qualified type.</para>
793 <term><option>-fwarn-missing-fields</option>:</term>
795 <indexterm><primary><option>-fwarn-missing-fields</option></primary></indexterm>
796 <indexterm><primary>missing fields, warning</primary></indexterm>
797 <indexterm><primary>fields, missing</primary></indexterm>
799 <para>This option is on by default, and warns you whenever
800 the construction of a labelled field constructor isn't
801 complete, missing initializers for one or more fields. While
802 not an error (the missing fields are initialised with
803 bottoms), it is often an indication of a programmer error.</para>
808 <term><option>-fwarn-missing-methods</option>:</term>
810 <indexterm><primary><option>-fwarn-missing-methods</option></primary></indexterm>
811 <indexterm><primary>missing methods, warning</primary></indexterm>
812 <indexterm><primary>methods, missing</primary></indexterm>
814 <para>This option is on by default, and warns you whenever
815 an instance declaration is missing one or more methods, and
816 the corresponding class declaration has no default
817 declaration for them.</para>
818 <para>The warning is suppressed if the method name
819 begins with an underscore. Here's an example where this is useful:
822 _simpleFn :: a -> String
823 complexFn :: a -> a -> String
824 complexFn x y = ... _simpleFn ...
826 The idea is that: (a) users of the class will only call <literal>complexFn</literal>;
827 never <literal>_simpleFn</literal>; and (b)
828 instance declarations can define either <literal>complexFn</literal> or <literal>_simpleFn</literal>.
834 <term><option>-fwarn-missing-signatures</option>:</term>
836 <indexterm><primary><option>-fwarn-missing-signatures</option></primary></indexterm>
837 <indexterm><primary>type signatures, missing</primary></indexterm>
839 <para>If you would like GHC to check that every top-level
840 function/value has a type signature, use the
841 <option>-fwarn-missing-signatures</option> option. This
842 option is off by default.</para>
847 <term><option>-fwarn-name-shadowing</option>:</term>
849 <indexterm><primary><option>-fwarn-name-shadowing</option></primary></indexterm>
850 <indexterm><primary>shadowing, warning</primary></indexterm>
852 <para>This option causes a warning to be emitted whenever an
853 inner-scope value has the same name as an outer-scope value,
854 i.e. the inner value shadows the outer one. This can catch
855 typographical errors that turn into hard-to-find bugs, e.g.,
856 in the inadvertent cyclic definition <literal>let x = ... x
857 ... in</literal>.</para>
859 <para>Consequently, this option does
860 <emphasis>will</emphasis> complain about cyclic recursive
866 <term><option>-fwarn-overlapping-patterns</option>:</term>
867 <indexterm><primary><option>-fwarn-overlapping-patterns</option></primary></indexterm>
868 <indexterm><primary>overlapping patterns, warning</primary></indexterm>
869 <indexterm><primary>patterns, overlapping</primary></indexterm>
871 <para>By default, the compiler will warn you if a set of
872 patterns are overlapping, i.e.,</para>
875 f :: String -> Int
881 <para>where the last pattern match in <function>f</function>
882 won't ever be reached, as the second pattern overlaps
883 it. More often than not, redundant patterns is a programmer
884 mistake/error, so this option is enabled by default.</para>
889 <term><option>-fwarn-simple-patterns</option>:</term>
891 <indexterm><primary><option>-fwarn-simple-patterns</option></primary>
893 <para>Causes the compiler to warn about lambda-bound
894 patterns that can fail, eg. <literal>\(x:xs)->...</literal>.
895 Normally, these aren't treated as incomplete patterns by
896 <option>-fwarn-incomplete-patterns</option>.</para>
897 <para>``Lambda-bound patterns'' includes all places where there is a single pattern,
898 including list comprehensions and do-notation. In these cases, a pattern-match
899 failure is quite legitimate, and triggers filtering (list comprehensions) or
900 the monad <literal>fail</literal> operation (monads). For example:
902 f :: [Maybe a] -> [a]
903 f xs = [y | Just y <- xs]
905 Switching on <option>-fwarn-simple-patterns</option> will elicit warnings about
906 these probably-innocent cases, which is why the flag is off by default. </para>
907 <para> The <literal>deriving( Read )</literal> mechanism produces monadic code with
908 pattern matches, so you will also get misleading warnings about the compiler-generated
909 code. (This is arguably a Bad Thing, but it's awkward to fix.)</para>
915 <term><option>-fwarn-type-defaults</option>:</term>
917 <indexterm><primary><option>-fwarn-type-defaults</option></primary></indexterm>
918 <indexterm><primary>defaulting mechanism, warning</primary></indexterm>
919 <para>Have the compiler warn/inform you where in your source
920 the Haskell defaulting mechanism for numeric types kicks
921 in. This is useful information when converting code from a
922 context that assumed one default into one with another,
923 e.g., the `default default' for Haskell 1.4 caused the
924 otherwise unconstrained value <constant>1</constant> to be
925 given the type <literal>Int</literal>, whereas Haskell 98
926 defaults it to <literal>Integer</literal>. This may lead to
927 differences in performance and behaviour, hence the
928 usefulness of being non-silent about this.</para>
930 <para>This warning is off by default.</para>
935 <term><option>-fwarn-unused-binds</option>:</term>
937 <indexterm><primary><option>-fwarn-unused-binds</option></primary></indexterm>
938 <indexterm><primary>unused binds, warning</primary></indexterm>
939 <indexterm><primary>binds, unused</primary></indexterm>
940 <para>Report any function definitions (and local bindings)
941 which are unused. For top-level functions, the warning is
942 only given if the binding is not exported.</para>
943 <para>A definition is regarded as "used" if (a) it is exported, or (b) it is
944 mentioned in the right hand side of another definition that is used, or (c) the
945 function it defines begins with an underscore. The last case provides a
946 way to suppress unused-binding warnings selectively. </para>
947 <para> Notice that a variable
948 is reported as unused even if it appears in the right-hand side of another
949 unused binding. </para>
954 <term><option>-fwarn-unused-imports</option>:</term>
956 <indexterm><primary><option>-fwarn-unused-imports</option></primary></indexterm>
957 <indexterm><primary>unused imports, warning</primary></indexterm>
958 <indexterm><primary>imports, unused</primary></indexterm>
960 <para>Report any modules that are explicitly imported but
961 never used. However, the form <literal>import M()</literal> is
962 never reported as an unused import, because it is a useful idiom
963 for importing instance declarations, which are anonymous in Haskell.</para>
968 <term><option>-fwarn-unused-matches</option>:</term>
970 <indexterm><primary><option>-fwarn-unused-matches</option></primary></indexterm>
971 <indexterm><primary>unused matches, warning</primary></indexterm>
972 <indexterm><primary>matches, unused</primary></indexterm>
974 <para>Report all unused variables which arise from pattern
975 matches, including patterns consisting of a single variable.
976 For instance <literal>f x y = []</literal> would report
977 <varname>x</varname> and <varname>y</varname> as unused. The
978 warning is suppressed if the variable name begins with an underscore, thus:
988 <para>If you're feeling really paranoid, the
989 <option>-dcore-lint</option>
990 option<indexterm><primary><option>-dcore-lint</option></primary></indexterm>
991 is a good choice. It turns on heavyweight intra-pass
992 sanity-checking within GHC. (It checks GHC's sanity, not
999 <sect1 id="options-optimise">
1000 <title>Optimisation (code improvement)</title>
1002 <indexterm><primary>optimisation</primary></indexterm>
1003 <indexterm><primary>improvement, code</primary></indexterm>
1005 <para>The <option>-O*</option> options specify convenient
1006 “packages” of optimisation flags; the
1007 <option>-f*</option> options described later on specify
1008 <emphasis>individual</emphasis> optimisations to be turned on/off;
1009 the <option>-m*</option> options specify
1010 <emphasis>machine-specific</emphasis> optimisations to be turned
1013 <sect2 id="optimise-pkgs">
1014 <title><option>-O*</option>: convenient “packages” of optimisation flags.</title>
1016 <para>There are <emphasis>many</emphasis> options that affect
1017 the quality of code produced by GHC. Most people only have a
1018 general goal, something like “Compile quickly” or
1019 “Make my program run like greased lightning.” The
1020 following “packages” of optimisations (or lack
1021 thereof) should suffice.</para>
1023 <para>Note that higher optimisation levels cause more
1024 cross-module optimisation to be performed, which can have an
1025 impact on how much of your program needs to be recompiled when
1026 you change something. This is one reaosn to stick to
1027 no-optimisation when developing code.</para>
1032 <term>No <option>-O*</option>-type option specified:</term>
1033 <indexterm><primary>-O* not specified</primary></indexterm>
1035 <para>This is taken to mean: “Please compile
1036 quickly; I'm not over-bothered about compiled-code
1037 quality.” So, for example: <command>ghc -c
1038 Foo.hs</command></para>
1043 <term><option>-O0</option>:</term>
1044 <indexterm><primary><option>-O0</option></primary></indexterm>
1046 <para>Means “turn off all optimisation”,
1047 reverting to the same settings as if no
1048 <option>-O</option> options had been specified. Saying
1049 <option>-O0</option> can be useful if
1050 eg. <command>make</command> has inserted a
1051 <option>-O</option> on the command line already.</para>
1056 <term><option>-O</option> or <option>-O1</option>:</term>
1057 <indexterm><primary>-O option</primary></indexterm>
1058 <indexterm><primary>-O1 option</primary></indexterm>
1059 <indexterm><primary>optimise</primary><secondary>normally</secondary></indexterm>
1061 <para>Means: “Generate good-quality code without
1062 taking too long about it.” Thus, for example:
1063 <command>ghc -c -O Main.lhs</command></para>
1065 <para><option>-O</option> currently also implies
1066 <option>-fvia-C</option>. This may change in the
1072 <term><option>-O2</option>:</term>
1073 <indexterm><primary>-O2 option</primary></indexterm>
1074 <indexterm><primary>optimise</primary><secondary>aggressively</secondary></indexterm>
1076 <para>Means: “Apply every non-dangerous
1077 optimisation, even if it means significantly longer
1078 compile times.”</para>
1080 <para>The avoided “dangerous” optimisations
1081 are those that can make runtime or space
1082 <emphasis>worse</emphasis> if you're unlucky. They are
1083 normally turned on or off individually.</para>
1085 <para>At the moment, <option>-O2</option> is
1086 <emphasis>unlikely</emphasis> to produce better code than
1087 <option>-O</option>.</para>
1092 <term><option>-Ofile <file></option>:</term>
1093 <indexterm><primary>-Ofile <file> option</primary></indexterm>
1094 <indexterm><primary>optimising, customised</primary></indexterm>
1096 <para>(NOTE: not supported since GHC 4.x. Please ask if
1097 you're interested in this.)</para>
1099 <para>For those who need <emphasis>absolute</emphasis>
1100 control over <emphasis>exactly</emphasis> what options are
1101 used (e.g., compiler writers, sometimes :-), a list of
1102 options can be put in a file and then slurped in with
1103 <option>-Ofile</option>.</para>
1105 <para>In that file, comments are of the
1106 <literal>#</literal>-to-end-of-line variety; blank
1107 lines and most whitespace is ignored.</para>
1109 <para>Please ask if you are baffled and would like an
1110 example of <option>-Ofile</option>!</para>
1115 <para>We don't use a <option>-O*</option> flag for day-to-day
1116 work. We use <option>-O</option> to get respectable speed;
1117 e.g., when we want to measure something. When we want to go for
1118 broke, we tend to use <option>-O2 -fvia-C</option> (and we go for
1119 lots of coffee breaks).</para>
1121 <para>The easiest way to see what <option>-O</option> (etc.)
1122 “really mean” is to run with <option>-v</option>,
1123 then stand back in amazement.</para>
1126 <sect2 id="options-f">
1127 <title><option>-f*</option>: platform-independent flags</title>
1129 <indexterm><primary>-f* options (GHC)</primary></indexterm>
1130 <indexterm><primary>-fno-* options (GHC)</primary></indexterm>
1132 <para>These flags turn on and off individual optimisations.
1133 They are normally set via the <option>-O</option> options
1134 described above, and as such, you shouldn't need to set any of
1135 them explicitly (indeed, doing so could lead to unexpected
1136 results). However, there are one or two that may be of
1141 <term><option>-fexcess-precision</option>:</term>
1143 <indexterm><primary><option>-fexcess-precision</option></primary></indexterm>
1144 <para>When this option is given, intermediate floating
1145 point values can have a <emphasis>greater</emphasis>
1146 precision/range than the final type. Generally this is a
1147 good thing, but some programs may rely on the exact
1149 <literal>Float</literal>/<literal>Double</literal> values
1150 and should not use this option for their compilation.</para>
1155 <term><option>-fignore-asserts</option>:</term>
1157 <indexterm><primary><option>-fignore-asserts</option></primary></indexterm>
1158 <para>Causes GHC to ignore uses of the function
1159 <literal>Exception.assert</literal> in source code (in
1160 other words, rewriting <literal>Exception.assert p
1161 e</literal> to <literal>e</literal> (see <xref
1162 linkend="sec-assertions"/>). This flag is turned on by
1163 <option>-O</option>.
1169 <term><option>-fno-strictness</option></term>
1170 <indexterm><primary><option>-fno-strictness</option></primary>
1173 <para>Turns off the strictness analyser; sometimes it eats
1174 too many cycles.</para>
1179 <term><option>-fno-cpr-analyse</option></term>
1180 <indexterm><primary><option>-fno-cpr-analyse</option></primary>
1183 <para>Turns off the CPR (constructed product result)
1184 analysis; it is somewhat experimental.</para>
1189 <term><option>-funbox-strict-fields</option>:</term>
1191 <indexterm><primary><option>-funbox-strict-fields</option></primary></indexterm>
1192 <indexterm><primary>strict constructor fields</primary></indexterm>
1193 <indexterm><primary>constructor fields, strict</primary></indexterm>
1195 <para>This option causes all constructor fields which are
1196 marked strict (i.e. “!”) to be unboxed or
1197 unpacked if possible. It is equivalent to adding an
1198 <literal>UNPACK</literal> pragma to every strict
1199 constructor field (see <xref
1200 linkend="unpack-pragma"/>).</para>
1202 <para>This option is a bit of a sledgehammer: it might
1203 sometimes make things worse. Selectively unboxing fields
1204 by using <literal>UNPACK</literal> pragmas might be
1210 <term><option>-funfolding-update-in-place<n></option></term>
1211 <indexterm><primary><option>-funfolding-update-in-place</option></primary></indexterm>
1213 <para>Switches on an experimental "optimisation".
1214 Switching it on makes the compiler a little keener to
1215 inline a function that returns a constructor, if the
1216 context is that of a thunk.
1220 If we inlined plusInt we might get an opportunity to use
1221 update-in-place for the thunk 'x'.</para>
1226 <term><option>-funfolding-creation-threshold<n></option>:</term>
1228 <indexterm><primary><option>-funfolding-creation-threshold</option></primary></indexterm>
1229 <indexterm><primary>inlining, controlling</primary></indexterm>
1230 <indexterm><primary>unfolding, controlling</primary></indexterm>
1232 <para>(Default: 45) Governs the maximum size that GHC will
1233 allow a function unfolding to be. (An unfolding has a
1234 “size” that reflects the cost in terms of
1235 “code bloat” of expanding that unfolding at
1236 at a call site. A bigger function would be assigned a
1237 bigger cost.) </para>
1239 <para> Consequences: (a) nothing larger than this will be
1240 inlined (unless it has an INLINE pragma); (b) nothing
1241 larger than this will be spewed into an interface
1245 <para> Increasing this figure is more likely to result in longer
1246 compile times than faster code. The next option is more
1252 <term><option>-funfolding-use-threshold<n></option>:</term>
1254 <indexterm><primary><option>-funfolding-use-threshold</option></primary></indexterm>
1255 <indexterm><primary>inlining, controlling</primary></indexterm>
1256 <indexterm><primary>unfolding, controlling</primary></indexterm>
1258 <para>(Default: 8) This is the magic cut-off figure for
1259 unfolding: below this size, a function definition will be
1260 unfolded at the call-site, any bigger and it won't. The
1261 size computed for a function depends on two things: the
1262 actual size of the expression minus any discounts that
1263 apply (see <option>-funfolding-con-discount</option>).</para>
1274 <sect1 id="sec-using-concurrent">
1275 <title>Using Concurrent Haskell</title>
1277 <indexterm><primary>Concurrent Haskell—use</primary></indexterm>
1280 GHC supports Concurrent Haskell by default, without requiring a
1281 special option or libraries compiled in a certain way. To get access
1282 to the support libraries for Concurrent Haskell, just import
1283 <literal>Control.Concurrent</literal> (details are in the accompanying
1284 library documentation).</para>
1287 RTS options are provided for modifying the behaviour of the threaded
1288 runtime system. See <xref linkend="parallel-rts-opts"/>.
1292 Concurrent Haskell is described in more detail in the documentation
1293 for the <literal>Control.Concurrent</literal> module.
1298 <sect1 id="sec-using-parallel">
1299 <title>Using parallel Haskell</title>
1302 <indexterm><primary>parallel Haskell—use</primary></indexterm>
1306 [You won't be able to execute parallel Haskell programs unless PVM3
1307 (parallel Virtual Machine, version 3) is installed at your site.]
1311 To compile a Haskell program for parallel execution under PVM, use the
1312 <option>-parallel</option> option,<indexterm><primary>-parallel
1313 option</primary></indexterm> both when compiling <emphasis>and
1314 linking</emphasis>. You will probably want to <literal>import
1315 parallel</literal> into your Haskell modules.
1319 To run your parallel program, once PVM is going, just invoke it
1320 “as normal”. The main extra RTS option is
1321 <option>-qp<n></option>, to say how many PVM
1322 “processors” your program to run on. (For more details of
1323 all relevant RTS options, please see <xref
1324 linkend="parallel-rts-opts"/>.)
1328 In truth, running parallel Haskell programs and getting information
1329 out of them (e.g., parallelism profiles) is a battle with the vagaries of
1330 PVM, detailed in the following sections.
1333 <sect2 id="pvm-dummies">
1334 <title>Dummy's guide to using PVM</title>
1337 <indexterm><primary>PVM, how to use</primary></indexterm>
1338 <indexterm><primary>parallel Haskell—PVM use</primary></indexterm>
1339 Before you can run a parallel program under PVM, you must set the
1340 required environment variables (PVM's idea, not ours); something like,
1341 probably in your <filename>.cshrc</filename> or equivalent:
1344 setenv PVM_ROOT /wherever/you/put/it
1345 setenv PVM_ARCH `$PVM_ROOT/lib/pvmgetarch`
1346 setenv PVM_DPATH $PVM_ROOT/lib/pvmd
1352 Creating and/or controlling your “parallel machine” is a purely-PVM
1353 business; nothing specific to parallel Haskell. The following paragraphs
1354 describe how to configure your parallel machine interactively.
1358 If you use parallel Haskell regularly on the same machine configuration it
1359 is a good idea to maintain a file with all machine names and to make the
1360 environment variable PVM_HOST_FILE point to this file. Then you can avoid
1361 the interactive operations described below by just saying
1369 You use the <command>pvm</command><indexterm><primary>pvm command</primary></indexterm> command to start PVM on your
1370 machine. You can then do various things to control/monitor your
1371 “parallel machine;” the most useful being:
1377 <colspec align="left"/>
1381 <entry><keycombo><keycap>Control</keycap><keycap>D</keycap></keycombo></entry>
1382 <entry>exit <command>pvm</command>, leaving it running</entry>
1386 <entry><command>halt</command></entry>
1387 <entry>kill off this “parallel machine” & exit</entry>
1391 <entry><command>add <host></command></entry>
1392 <entry>add <command><host></command> as a processor</entry>
1396 <entry><command>delete <host></command></entry>
1397 <entry>delete <command><host></command></entry>
1401 <entry><command>reset</command></entry>
1402 <entry>kill what's going, but leave PVM up</entry>
1406 <entry><command>conf</command></entry>
1407 <entry>list the current configuration</entry>
1411 <entry><command>ps</command></entry>
1412 <entry>report processes' status</entry>
1416 <entry><command>pstat <pid></command></entry>
1417 <entry>status of a particular process</entry>
1426 The PVM documentation can tell you much, much more about <command>pvm</command>!
1431 <sect2 id="par-profiles">
1432 <title>parallelism profiles</title>
1435 <indexterm><primary>parallelism profiles</primary></indexterm>
1436 <indexterm><primary>profiles, parallelism</primary></indexterm>
1437 <indexterm><primary>visualisation tools</primary></indexterm>
1441 With parallel Haskell programs, we usually don't care about the
1442 results—only with “how parallel” it was! We want pretty pictures.
1446 parallelism profiles (à la <command>hbcpp</command>) can be generated with the
1447 <option>-qP</option><indexterm><primary>-qP RTS option (concurrent, parallel)</primary></indexterm> RTS option. The
1448 per-processor profiling info is dumped into files named
1449 <filename><full-path><program>.gr</filename>. These are then munged into a PostScript picture,
1450 which you can then display. For example, to run your program
1451 <filename>a.out</filename> on 8 processors, then view the parallelism profile, do:
1457 <prompt>$</prompt> ./a.out +RTS -qP -qp8
1458 <prompt>$</prompt> grs2gr *.???.gr > temp.gr # combine the 8 .gr files into one
1459 <prompt>$</prompt> gr2ps -O temp.gr # cvt to .ps; output in temp.ps
1460 <prompt>$</prompt> ghostview -seascape temp.ps # look at it!
1466 The scripts for processing the parallelism profiles are distributed
1467 in <filename>ghc/utils/parallel/</filename>.
1473 <title>Other useful info about running parallel programs</title>
1476 The “garbage-collection statistics” RTS options can be useful for
1477 seeing what parallel programs are doing. If you do either
1478 <option>+RTS -Sstderr</option><indexterm><primary>-Sstderr RTS option</primary></indexterm> or <option>+RTS -sstderr</option>, then
1479 you'll get mutator, garbage-collection, etc., times on standard
1480 error. The standard error of all PE's other than the `main thread'
1481 appears in <filename>/tmp/pvml.nnn</filename>, courtesy of PVM.
1485 Whether doing <option>+RTS -Sstderr</option> or not, a handy way to watch
1486 what's happening overall is: <command>tail -f /tmp/pvml.nnn</command>.
1491 <sect2 id="parallel-rts-opts">
1492 <title>RTS options for Concurrent/parallel Haskell
1496 <indexterm><primary>RTS options, concurrent</primary></indexterm>
1497 <indexterm><primary>RTS options, parallel</primary></indexterm>
1498 <indexterm><primary>Concurrent Haskell—RTS options</primary></indexterm>
1499 <indexterm><primary>parallel Haskell—RTS options</primary></indexterm>
1503 Besides the usual runtime system (RTS) options
1504 (<xref linkend="runtime-control"/>), there are a few options particularly
1505 for concurrent/parallel execution.
1512 <term><option>-qp<N></option>:</term>
1515 <indexterm><primary>-qp<N> RTS option</primary></indexterm>
1516 (paraLLEL ONLY) Use <literal><N></literal> PVM processors to run this program;
1522 <term><option>-C[<us>]</option>:</term>
1525 <indexterm><primary>-C<us> RTS option</primary></indexterm> Sets
1526 the context switch interval to <literal><s></literal> seconds.
1527 A context switch will occur at the next heap block allocation after
1528 the timer expires (a heap block allocation occurs every 4k of
1529 allocation). With <option>-C0</option> or <option>-C</option>,
1530 context switches will occur as often as possible (at every heap block
1531 allocation). By default, context switches occur every 20ms
1532 milliseconds. Note that GHC's internal timer ticks every 20ms, and
1533 the context switch timer is always a multiple of this timer, so 20ms
1534 is the maximum granularity available for timed context switches.
1539 <term><option>-q[v]</option>:</term>
1542 <indexterm><primary>-q RTS option</primary></indexterm>
1543 (paraLLEL ONLY) Produce a quasi-parallel profile of thread activity,
1544 in the file <filename><program>.qp</filename>. In the style of <command>hbcpp</command>, this profile
1545 records the movement of threads between the green (runnable) and red
1546 (blocked) queues. If you specify the verbose suboption (<option>-qv</option>), the
1547 green queue is split into green (for the currently running thread
1548 only) and amber (for other runnable threads). We do not recommend
1549 that you use the verbose suboption if you are planning to use the
1550 <command>hbcpp</command> profiling tools or if you are context switching at every heap
1551 check (with <option>-C</option>).
1557 <term><option>-qt<num></option>:</term>
1560 <indexterm><primary>-qt<num> RTS option</primary></indexterm>
1561 (paraLLEL ONLY) Limit the thread pool size, i.e. the number of concurrent
1562 threads per processor to <literal><num></literal>. The default is
1563 32. Each thread requires slightly over 1K <emphasis>words</emphasis> in
1564 the heap for thread state and stack objects. (For 32-bit machines, this
1565 translates to 4K bytes, and for 64-bit machines, 8K bytes.)
1571 <term><option>-d</option>:</term>
1574 <indexterm><primary>-d RTS option (parallel)</primary></indexterm>
1575 (paraLLEL ONLY) Turn on debugging. It pops up one xterm (or GDB, or
1576 something…) per PVM processor. We use the standard <command>debugger</command>
1577 script that comes with PVM3, but we sometimes meddle with the
1578 <command>debugger2</command> script. We include ours in the GHC distribution,
1579 in <filename>ghc/utils/pvm/</filename>.
1585 <term><option>-qe<num></option>:</term>
1588 <indexterm><primary>-qe<num> RTS option
1589 (parallel)</primary></indexterm> (paraLLEL ONLY) Limit the spark pool size
1590 i.e. the number of pending sparks per processor to
1591 <literal><num></literal>. The default is 100. A larger number may be
1592 appropriate if your program generates large amounts of parallelism
1598 <term><option>-qQ<num></option>:</term>
1601 <indexterm><primary>-qQ<num> RTS option (parallel)</primary></indexterm>
1602 (paraLLEL ONLY) Set the size of packets transmitted between processors
1603 to <literal><num></literal>. The default is 1024 words. A larger number may be
1604 appropriate if your machine has a high communication cost relative to
1610 <term><option>-qh<num></option>:</term>
1613 <indexterm><primary>-qh<num> RTS option (parallel)</primary></indexterm>
1614 (paraLLEL ONLY) Select a packing scheme. Set the number of non-root thunks to pack in one packet to
1615 <num>-1 (0 means infinity). By default GUM uses full-subgraph
1616 packing, i.e. the entire subgraph with the requested closure as root is
1617 transmitted (provided it fits into one packet). Choosing a smaller value
1618 reduces the amount of pre-fetching of work done in GUM. This can be
1619 advantageous for improving data locality but it can also worsen the balance
1620 of the load in the system.
1625 <term><option>-qg<num></option>:</term>
1628 <indexterm><primary>-qg<num> RTS option
1629 (parallel)</primary></indexterm> (paraLLEL ONLY) Select a globalisation
1630 scheme. This option affects the
1631 generation of global addresses when transferring data. Global addresses are
1632 globally unique identifiers required to maintain sharing in the distributed
1633 graph structure. Currently this is a binary option. With <num>=0 full globalisation is used
1634 (default). This means a global address is generated for every closure that
1635 is transmitted. With <num>=1 a thunk-only globalisation scheme is
1636 used, which generated global address only for thunks. The latter case may
1637 lose sharing of data but has a reduced overhead in packing graph structures
1638 and maintaining internal tables of global addresses.
1649 <sect1 id="options-platform">
1650 <title>Platform-specific Flags</title>
1652 <indexterm><primary>-m* options</primary></indexterm>
1653 <indexterm><primary>platform-specific options</primary></indexterm>
1654 <indexterm><primary>machine-specific options</primary></indexterm>
1656 <para>Some flags only make sense for particular target
1662 <term><option>-mv8</option>:</term>
1664 <para>(SPARC machines)<indexterm><primary>-mv8 option (SPARC
1665 only)</primary></indexterm> Means to pass the like-named
1666 option to GCC; it says to use the Version 8 SPARC
1667 instructions, notably integer multiply and divide. The
1668 similiar <option>-m*</option> GCC options for SPARC also
1669 work, actually.</para>
1674 <term><option>-monly-[32]-regs</option>:</term>
1676 <para>(iX86 machines)<indexterm><primary>-monly-N-regs
1677 option (iX86 only)</primary></indexterm> GHC tries to
1678 “steal” four registers from GCC, for performance
1679 reasons; it almost always works. However, when GCC is
1680 compiling some modules with four stolen registers, it will
1681 crash, probably saying:
1684 Foo.hc:533: fixed or forbidden register was spilled.
1685 This may be due to a compiler bug or to impossible asm
1686 statements or clauses.
1689 Just give some registers back with
1690 <option>-monly-N-regs</option>. Try `3' first, then `2'.
1691 If `2' doesn't work, please report the bug to us.</para>
1700 <sect1 id="ext-core">
1701 <title>Generating and compiling External Core Files</title>
1703 <indexterm><primary>intermediate code generation</primary></indexterm>
1705 <para>GHC can dump its optimized intermediate code (said to be in “Core” format)
1706 to a file as a side-effect of compilation. Core files, which are given the suffix
1707 <filename>.hcr</filename>, can be read and processed by non-GHC back-end
1708 tools. The Core format is formally described in <ulink url="http://www.haskell.org/ghc/docs/papers/core.ps.gz">
1709 <citetitle>An External Representation for the GHC Core Language</citetitle></ulink>,
1710 and sample tools (in Haskell)
1711 for manipulating Core files are available in the GHC source distribution
1712 directory <literal>/fptools/ghc/utils/ext-core</literal>.
1713 Note that the format of <literal>.hcr</literal>
1714 files is <emphasis>different</emphasis> (though similar) to the Core output format generated
1715 for debugging purposes (<xref linkend="options-debugging"/>).</para>
1717 <para>The Core format natively supports notes which you can add to
1718 your source code using the <literal>CORE</literal> pragma (see <xref
1719 linkend="pragmas"/>).</para>
1724 <term><option>-fext-core</option></term>
1726 <primary><option>-fext-core</option></primary>
1729 <para>Generate <literal>.hcr</literal> files.</para>
1735 <para>GHC can also read in External Core files as source; just give the <literal>.hcr</literal> file on
1736 the command line, instead of the <literal>.hs</literal> or <literal>.lhs</literal> Haskell source.
1737 A current infelicity is that you need to give teh <literal>-fglasgow-exts</literal> flag too, because
1738 ordinary Haskell 98, when translated to External Core, uses things like rank-2 types.</para>
1747 ;;; Local Variables: ***
1749 ;;; sgml-parent-document: ("users_guide.sgml" "book" "chapter") ***