1 <?xml version="1.0" encoding="iso-8859-1"?>
2 <chapter id="using-ghc">
3 <title>Using GHC</title>
5 <indexterm><primary>GHC, using</primary></indexterm>
6 <indexterm><primary>using GHC</primary></indexterm>
9 <title>Options overview</title>
11 <para>GHC's behaviour is controlled by
12 <firstterm>options</firstterm>, which for historical reasons are
13 also sometimes referred to as command-line flags or arguments.
14 Options can be specified in three ways:</para>
17 <title>command-line arguments</title>
19 <indexterm><primary>structure, command-line</primary></indexterm>
20 <indexterm><primary>command-line</primary><secondary>arguments</secondary></indexterm>
21 <indexterm><primary>arguments</primary><secondary>command-line</secondary></indexterm>
23 <para>An invocation of GHC takes the following form:</para>
29 <para>command-line arguments are either options or file names.</para>
31 <para>command-line options begin with <literal>-</literal>.
32 They may <emphasis>not</emphasis> be grouped:
33 <option>-vO</option> is different from <option>-v -O</option>.
34 Options need not precede filenames: e.g., <literal>ghc *.o -o
35 foo</literal>. All options are processed and then applied to
36 all files; you cannot, for example, invoke <literal>ghc -c -O1
37 Foo.hs -O2 Bar.hs</literal> to apply different optimisation
38 levels to the files <filename>Foo.hs</filename> and
39 <filename>Bar.hs</filename>.</para>
42 <sect2 id="source-file-options">
43 <title>command line options in source files</title>
45 <indexterm><primary>source-file options</primary></indexterm>
47 <para>Sometimes it is useful to make the connection between a
48 source file and the command-line options it requires quite
49 tight. For instance, if a Haskell source file uses GHC
50 extensions, it will always need to be compiled with the
51 <option>-fglasgow-exts</option> option. Rather than maintaining
52 the list of per-file options in a <filename>Makefile</filename>,
53 it is possible to do this directly in the source file using the
54 <literal>OPTIONS_GHC</literal> pragma <indexterm><primary>OPTIONS_GHC
55 pragma</primary></indexterm>:</para>
58 {-# OPTIONS_GHC -fglasgow-exts #-}
63 <para><literal>OPTIONS_GHC</literal> pragmas are only looked for at
64 the top of your source files, upto the first
65 (non-literate,non-empty) line not containing
66 <literal>OPTIONS_GHC</literal>. Multiple <literal>OPTIONS_GHC</literal>
67 pragmas are recognised. Do not put comments before, or on the same line
68 as, the <literal>OPTIONS_GHC</literal> pragma.</para>
70 <para>Note that your command shell does not
71 get to the source file options, they are just included literally
72 in the array of command-line arguments the compiler
73 maintains internally, so you'll be desperately disappointed if
74 you try to glob etc. inside <literal>OPTIONS_GHC</literal>.</para>
76 <para>NOTE: the contents of OPTIONS_GHC are prepended to the
77 command-line options, so you <emphasis>do</emphasis> have the
78 ability to override OPTIONS_GHC settings via the command
81 <para>It is not recommended to move all the contents of your
82 Makefiles into your source files, but in some circumstances, the
83 <literal>OPTIONS_GHC</literal> pragma is the Right Thing. (If you
84 use <option>-keep-hc-file-too</option> and have OPTION flags in
85 your module, the OPTIONS_GHC will get put into the generated .hc
90 <title>Setting options in GHCi</title>
92 <para>Options may also be modified from within GHCi, using the
93 <literal>:set</literal> command. See <xref linkend="ghci-set"/>
94 for more details.</para>
98 <sect1 id="static-dynamic-flags">
99 <title>Static, Dynamic, and Mode options</title>
100 <indexterm><primary>static</primary><secondary>options</secondary>
102 <indexterm><primary>dynamic</primary><secondary>options</secondary>
104 <indexterm><primary>mode</primary><secondary>options</secondary>
107 <para>Each of GHC's command line options is classified as either
108 <firstterm>static</firstterm> or <firstterm>dynamic</firstterm> or
109 <firstterm>mode</firstterm>:</para>
113 <term>Mode flags</term>
115 <para>For example, <option>--make</option> or <option>-E</option>.
116 There may be only a single mode flag on the command line. The
117 available modes are listed in <xref linkend="modes"/>.</para>
121 <term>Dynamic Flags</term>
123 <para>Most non-mode flags fall into this category. A dynamic flag
124 may be used on the command line, in a
125 <literal>GHC_OPTIONS</literal> pragma in a source file, or set
126 using <literal>:set</literal> in GHCi.</para>
130 <term>Static Flags</term>
132 <para>A few flags are "static", which means they can only be used on
133 the command-line, and remain in force over the entire GHC/GHCi
139 <para>The flag reference tables (<xref
140 linkend="flag-reference"/>) lists the status of each flag.</para>
143 <sect1 id="file-suffixes">
144 <title>Meaningful file suffixes</title>
146 <indexterm><primary>suffixes, file</primary></indexterm>
147 <indexterm><primary>file suffixes for GHC</primary></indexterm>
149 <para>File names with “meaningful” suffixes (e.g.,
150 <filename>.lhs</filename> or <filename>.o</filename>) cause the
151 “right thing” to happen to those files.</para>
156 <term><filename>.hs</filename></term>
158 <para>A Haskell module.</para>
164 <filename>.lhs</filename>
165 <indexterm><primary><literal>lhs</literal> suffix</primary></indexterm>
168 <para>A “literate Haskell” module.</para>
173 <term><filename>.hi</filename></term>
175 <para>A Haskell interface file, probably
176 compiler-generated.</para>
181 <term><filename>.hc</filename></term>
183 <para>Intermediate C file produced by the Haskell
189 <term><filename>.c</filename></term>
191 <para>A C file not produced by the Haskell
197 <term><filename>.s</filename></term>
199 <para>An assembly-language source file, usually produced by
205 <term><filename>.o</filename></term>
207 <para>An object file, produced by an assembler.</para>
212 <para>Files with other suffixes (or without suffixes) are passed
213 straight to the linker.</para>
218 <title>Modes of operation</title>
220 <para>GHC's behaviour is firstly controlled by a mode flag. Only
221 one of these flags may be given, but it does not necessarily need
222 to be the first option on the command-line. The available modes
228 <cmdsynopsis><command>ghc</command>
229 <arg choice='plain'>––interactive</arg>
231 <indexterm><primary>interactive mode</primary></indexterm>
232 <indexterm><primary>ghci</primary></indexterm>
235 <para>Interactive mode, which is also available as
236 <command>ghci</command>. Interactive mode is described in
237 more detail in <xref linkend="ghci"/>.</para>
243 <cmdsynopsis><command>ghc</command>
244 <arg choice='plain'>––make</arg>
246 <indexterm><primary>make mode</primary></indexterm>
247 <indexterm><primary><option>––make</option></primary></indexterm>
250 <para>In this mode, GHC will build a multi-module Haskell
251 program automatically, figuring out dependencies for itself.
252 If you have a straightforward Haskell program, this is
253 likely to be much easier, and faster, than using
254 <command>make</command>. Make mode is described in <xref
255 linkend="make-mode"/>.</para>
261 <cmdsynopsis><command>ghc</command>
262 <arg choice='plain'>–e</arg> <arg choice='plain'><replaceable>expr</replaceable></arg>
264 <indexterm><primary>eval mode</primary></indexterm>
267 <para>Expression-evaluation mode. This is very similar to
268 interactive mode, except that there is a single expression
269 to evaluate (<replaceable>expr</replaceable>) which is given
270 on the command line. See <xref linkend="eval-mode"/> for
278 <command>ghc</command>
286 <indexterm><primary><option>-E</option></primary></indexterm>
287 <indexterm><primary><option>-C</option></primary></indexterm>
288 <indexterm><primary><option>-S</option></primary></indexterm>
289 <indexterm><primary><option>-c</option></primary></indexterm>
292 <para>This is the traditional batch-compiler mode, in which
293 GHC can compile source files one at a time, or link objects
294 together into an executable. This mode also applies if
295 there is no other mode flag specified on the command line,
296 in which case it means that the specified files should be
297 compiled and then linked to form a program. See <xref
298 linkend="options-order"/>.</para>
305 <command>ghc</command>
306 <arg choice='plain'>–M</arg>
308 <indexterm><primary>dependency-generation mode</primary></indexterm>
311 <para>Dependency-generation mode. In this mode, GHC can be
312 used to generate dependency information suitable for use in
313 a <literal>Makefile</literal>. See <xref
314 linkend="sec-makefile-dependencies"/>.</para>
321 <command>ghc</command>
322 <arg choice='plain'>––mk-dll</arg>
324 <indexterm><primary>dependency-generation mode</primary></indexterm>
327 <para>DLL-creation mode (Windows only). See <xref
328 linkend="win32-dlls-create"/>.</para>
333 <sect2 id="make-mode">
334 <title>Using <command>ghc</command> <option>––make</option></title>
335 <indexterm><primary><option>––make</option></primary></indexterm>
336 <indexterm><primary>separate compilation</primary></indexterm>
338 <para>When given the <option>––make</option> option,
339 GHC will build a multi-module Haskell program by following
340 dependencies from a single root module (usually
341 <literal>Main</literal>). For example, if your
342 <literal>Main</literal> module is in a file called
343 <filename>Main.hs</filename>, you could compile and link the
344 program like this:</para>
347 ghc ––make Main.hs
350 <para>The command line may contain any number of source file
351 names or module names; GHC will figure out all the modules in
352 the program by following the imports from these initial modules.
353 It will then attempt to compile each module which is out of
354 date, and finally if there is a <literal>Main</literal> module,
355 the program will also be linked into an executable.</para>
357 <para>The main advantages to using <literal>ghc
358 ––make</literal> over traditional
359 <literal>Makefile</literal>s are:</para>
363 <para>GHC doesn't have to be restarted for each compilation,
364 which means it can cache information between compilations.
365 Compiling a multi-module program with <literal>ghc
366 ––make</literal> can be up to twice as fast as
367 running <literal>ghc</literal> individually on each source
371 <para>You don't have to write a <literal>Makefile</literal>.</para>
372 <indexterm><primary><literal>Makefile</literal>s</primary><secondary>avoiding</secondary></indexterm>
375 <para>GHC re-calculates the dependencies each time it is
376 invoked, so the dependencies never get out of sync with the
381 <para>Any of the command-line options described in the rest of
382 this chapter can be used with
383 <option>––make</option>, but note that any options
384 you give on the command line will apply to all the source files
385 compiled, so if you want any options to apply to a single source
386 file only, you'll need to use an <literal>OPTIONS_GHC</literal>
387 pragma (see <xref linkend="source-file-options"/>).</para>
389 <para>If the program needs to be linked with additional objects
390 (say, some auxiliary C code), then the object files can be
391 given on the command line and GHC will include them when linking
392 the executable.</para>
394 <para>Note that GHC can only follow dependencies if it has the
395 source file available, so if your program includes a module for
396 which there is no source file, even if you have an object and an
397 interface file for the module, then GHC will complain. The
398 exception to this rule is for package modules, which may or may
399 not have source files.</para>
401 <para>The source files for the program don't all need to be in
402 the same directory; the <option>-i</option> option can be used
403 to add directories to the search path (see <xref
404 linkend="search-path"/>).</para>
407 <sect2 id="eval-mode">
408 <title>Expression evaluation mode</title>
410 <para>This mode is very similar to interactive mode, except that
411 there is a single expression to evaluate which is specified on
412 the command line as an argument to the <option>-e</option>
416 ghc -e <replaceable>expr</replaceable>
419 <para>Haskell source files may be named on the command line, and
420 they will be loaded exactly as in interactive mode. The
421 expression is evaluated in the context of the loaded
424 <para>For example, to load and run a Haskell program containing
425 a module <literal>Main</literal>, we might say</para>
428 ghc -e Main.main Main.hs
431 <para>or we can just use this mode to evaluate expressions in
432 the context of the <literal>Prelude</literal>:</para>
435 $ ghc -e "interact (unlines.map reverse.lines)"
441 <sect2 id="options-order">
442 <title>Batch compiler mode</title>
444 <para>In <emphasis>batch mode</emphasis>, GHC will compile one or more source files
445 given on the command line.</para>
447 <para>The first phase to run is determined by each input-file
448 suffix, and the last phase is determined by a flag. If no
449 relevant flag is present, then go all the way through linking.
450 This table summarises:</para>
454 <colspec align="left"/>
455 <colspec align="left"/>
456 <colspec align="left"/>
457 <colspec align="left"/>
461 <entry>Phase of the compilation system</entry>
462 <entry>Suffix saying “start here”</entry>
463 <entry>Flag saying “stop after”</entry>
464 <entry>(suffix of) output file</entry>
469 <entry>literate pre-processor</entry>
470 <entry><literal>.lhs</literal></entry>
472 <entry><literal>.hs</literal></entry>
476 <entry>C pre-processor (opt.) </entry>
477 <entry><literal>.hs</literal> (with
478 <option>-cpp</option>)</entry>
479 <entry><option>-E</option></entry>
480 <entry><literal>.hspp</literal></entry>
484 <entry>Haskell compiler</entry>
485 <entry><literal>.hs</literal></entry>
486 <entry><option>-C</option>, <option>-S</option></entry>
487 <entry><literal>.hc</literal>, <literal>.s</literal></entry>
491 <entry>C compiler (opt.)</entry>
492 <entry><literal>.hc</literal> or <literal>.c</literal></entry>
493 <entry><option>-S</option></entry>
494 <entry><literal>.s</literal></entry>
498 <entry>assembler</entry>
499 <entry><literal>.s</literal></entry>
500 <entry><option>-c</option></entry>
501 <entry><literal>.o</literal></entry>
505 <entry>linker</entry>
506 <entry><replaceable>other</replaceable></entry>
508 <entry><filename>a.out</filename></entry>
514 <indexterm><primary><option>-C</option></primary></indexterm>
515 <indexterm><primary><option>-E</option></primary></indexterm>
516 <indexterm><primary><option>-S</option></primary></indexterm>
517 <indexterm><primary><option>-c</option></primary></indexterm>
519 <para>Thus, a common invocation would be: </para>
522 ghc -c Foo.hs</screen>
524 <para>to compile the Haskell source file
525 <filename>Foo.hs</filename> to an object file
526 <filename>Foo.o</filename>.</para>
528 <para>Note: What the Haskell compiler proper produces depends on
529 whether a native-code generator<indexterm><primary>native-code
530 generator</primary></indexterm> is used (producing assembly
531 language) or not (producing C). See <xref
532 linkend="options-codegen"/> for more details.</para>
534 <para>Note: C pre-processing is optional, the
535 <option>-cpp</option><indexterm><primary><option>-cpp</option></primary></indexterm>
536 flag turns it on. See <xref linkend="c-pre-processor"/> for more
539 <para>Note: The option <option>-E</option><indexterm><primary>-E
540 option</primary></indexterm> runs just the pre-processing passes
541 of the compiler, dumping the result in a file. Note that this
542 differs from the previous behaviour of dumping the file to
543 standard output.</para>
545 <sect3 id="overriding-suffixes">
546 <title>Overriding the default behaviour for a file</title>
548 <para>As described above, the way in which a file is processed by GHC
549 depends on its suffix. This behaviour can be overriden using the
550 <option>-x</option> option:</para>
554 <term><option>-x</option> <replaceable>suffix</replaceable>
555 <indexterm><primary><option>-x</option></primary>
558 <para>Causes all files following this option on the command
559 line to be processed as if they had the suffix
560 <replaceable>suffix</replaceable>. For example, to compile a
561 Haskell module in the file <literal>M.my-hs</literal>,
562 use <literal>ghc -c -x hs M.my-hs</literal>.</para>
571 <sect1 id="options-help">
572 <title>Help and verbosity options</title>
574 <indexterm><primary>help options</primary></indexterm>
575 <indexterm><primary>verbosity options</primary></indexterm>
580 <option>––help</option>
581 <indexterm><primary><option>––help</option></primary></indexterm>
585 <indexterm><primary><option>-?</option></primary></indexterm>
588 <para>Cause GHC to spew a long usage message to standard
589 output and then exit.</para>
596 <indexterm><primary><option>-v</option></primary></indexterm>
599 <para>The <option>-v</option> option makes GHC
600 <emphasis>verbose</emphasis>: it reports its version number
601 and shows (on stderr) exactly how it invokes each phase of
602 the compilation system. Moreover, it passes the
603 <option>-v</option> flag to most phases; each reports its
604 version number (and possibly some other information).</para>
606 <para>Please, oh please, use the <option>-v</option> option
607 when reporting bugs! Knowing that you ran the right bits in
608 the right order is always the first thing we want to
615 <option>-v</option><replaceable>n</replaceable>
616 <indexterm><primary><option>-v</option></primary></indexterm>
619 <para>To provide more control over the compiler's verbosity,
620 the <option>-v</option> flag takes an optional numeric
621 argument. Specifying <option>-v</option> on its own is
622 equivalent to <option>-v3</option>, and the other levels
623 have the following meanings:</para>
627 <term><option>-v0</option></term>
629 <para>Disable all non-essential messages (this is the
635 <term><option>-v1</option></term>
637 <para>Minimal verbosity: print one line per
638 compilation (this is the default when
639 <option>––make</option> or
640 <option>––interactive</option> is on).</para>
645 <term><option>-v2</option></term>
647 <para>Print the name of each compilation phase as it
648 is executed. (equivalent to
649 <option>-dshow-passes</option>).</para>
654 <term><option>-v3</option></term>
656 <para>The same as <option>-v2</option>, except that in
657 addition the full command line (if appropriate) for
658 each compilation phase is also printed.</para>
663 <term><option>-v4</option></term>
665 <para>The same as <option>-v3</option> except that the
666 intermediate program representation after each
667 compilation phase is also printed (excluding
668 preprocessed and C/assembly files).</para>
678 <indexterm><primary><option>-V</option></primary></indexterm>
681 <option>––version</option>
682 <indexterm><primary><option>––version</option></primary></indexterm>
685 <para>Print a one-line string including GHC's version number.</para>
691 <option>––numeric-version</option>
692 <indexterm><primary><option>––numeric-version</option></primary></indexterm>
695 <para>Print GHC's numeric version number only.</para>
701 <option>––print-libdir</option>
702 <indexterm><primary><option>––print-libdir</option></primary></indexterm>
705 <para>Print the path to GHC's library directory. This is
706 the top of the directory tree containing GHC's libraries,
707 interfaces, and include files (usually something like
708 <literal>/usr/local/lib/ghc-5.04</literal> on Unix). This
710 <literal>$libdir</literal><indexterm><primary><literal>libdir</literal></primary>
711 </indexterm>in the package configuration file (see <xref
712 linkend="packages"/>).</para>
717 <term><option>-ferror-spans</option>
718 <indexterm><primary><option>-ferror-spans</option></primary>
722 <para>Causes GHC to emit the full source span of the
723 syntactic entity relating to an error message. Normally, GHC
724 emits the source location of the start of the syntactic
727 <para>For example:</para>
729 <screen>test.hs:3:6: parse error on input `where'</screen>
731 <para>becomes:</para>
733 <screen>test296.hs:3:6-10: parse error on input `where'</screen>
735 <para>And multi-line spans are possible too:</para>
737 <screen>test.hs:(5,4)-(6,7):
738 Conflicting definitions for `a'
739 Bound at: test.hs:5:4
741 In the binding group for: a, b, a</screen>
743 <para>Note that line numbers start counting at one, but
744 column numbers start at zero. This choice was made to
745 follow existing convention (i.e. this is how Emacs does
751 <term><option>-Rghc-timing</option>
752 <indexterm><primary><option>-Rghc-timing</option></primary></indexterm>
755 <para>Prints a one-line summary of timing statistics for the
756 GHC run. This option is equivalent to
757 <literal>+RTS -tstderr</literal>, see <xref
758 linkend="rts-options-gc" />.
767 <sect1 id="options-sanity">
768 <title>Warnings and sanity-checking</title>
770 <indexterm><primary>sanity-checking options</primary></indexterm>
771 <indexterm><primary>warnings</primary></indexterm>
774 <para>GHC has a number of options that select which types of
775 non-fatal error messages, otherwise known as warnings, can be
776 generated during compilation. By default, you get a standard set
777 of warnings which are generally likely to indicate bugs in your
779 <option>-fwarn-overlapping-patterns</option>,
780 <option>-fwarn-deprecations</option>,
781 <option>-fwarn-duplicate-exports</option>,
782 <option>-fwarn-missing-fields</option>, and
783 <option>-fwarn-missing-methods</option>. The following flags are
784 simple ways to select standard “packages” of warnings:
790 <term><option>-W</option>:</term>
792 <indexterm><primary>-W option</primary></indexterm>
793 <para>Provides the standard warnings plus
794 <option>-fwarn-incomplete-patterns</option>,
795 <option>-fwarn-unused-matches</option>,
796 <option>-fwarn-unused-imports</option>,
797 <option>-fwarn-misc</option>, and
798 <option>-fwarn-unused-binds</option>.</para>
803 <term><option>-w</option>:</term>
805 <indexterm><primary><option>-w</option></primary></indexterm>
806 <para>Turns off all warnings, including the standard ones.</para>
811 <term><option>-Wall</option>:</term>
813 <indexterm><primary><option>-Wall</option></primary></indexterm>
814 <para>Turns on all warning options.</para>
819 <term><option>-Werror</option>:</term>
821 <indexterm><primary><option>-Werror</option></primary></indexterm>
822 <para>Makes any warning into a fatal error. Useful so that you don't
823 miss warnings when doing batch compilation. </para>
829 <para>The full set of warning options is described below. To turn
830 off any warning, simply give the corresponding
831 <option>-fno-warn-...</option> option on the command line.</para>
836 <term><option>-fwarn-deprecations</option>:</term>
838 <indexterm><primary><option>-fwarn-deprecations</option></primary>
840 <indexterm><primary>deprecations</primary></indexterm>
841 <para>Causes a warning to be emitted when a deprecated
842 function or type is used. Entities can be marked as
843 deprecated using a pragma, see <xref
844 linkend="deprecated-pragma"/>.</para>
849 <term><option>-fwarn-duplicate-exports</option>:</term>
851 <indexterm><primary><option>-fwarn-duplicate-exports</option></primary></indexterm>
852 <indexterm><primary>duplicate exports, warning</primary></indexterm>
853 <indexterm><primary>export lists, duplicates</primary></indexterm>
855 <para>Have the compiler warn about duplicate entries in
856 export lists. This is useful information if you maintain
857 large export lists, and want to avoid the continued export
858 of a definition after you've deleted (one) mention of it in
859 the export list.</para>
861 <para>This option is on by default.</para>
866 <term><option>-fwarn-hi-shadowing</option>:</term>
868 <indexterm><primary><option>-fwarn-hi-shadowing</option></primary></indexterm>
869 <indexterm><primary>shadowing</primary>
870 <secondary>interface files</secondary></indexterm>
872 <para>Causes the compiler to emit a warning when a module or
873 interface file in the current directory is shadowing one
874 with the same module name in a library or other
880 <term><option>-fwarn-incomplete-patterns</option>:</term>
882 <indexterm><primary><option>-fwarn-incomplete-patterns</option></primary></indexterm>
883 <indexterm><primary>incomplete patterns, warning</primary></indexterm>
884 <indexterm><primary>patterns, incomplete</primary></indexterm>
886 <para>Similarly for incomplete patterns, the function
887 <function>g</function> below will fail when applied to
888 non-empty lists, so the compiler will emit a warning about
889 this when <option>-fwarn-incomplete-patterns</option> is
896 <para>This option isn't enabled be default because it can be
897 a bit noisy, and it doesn't always indicate a bug in the
898 program. However, it's generally considered good practice
899 to cover all the cases in your functions.</para>
904 <term><option>-fwarn-incomplete-record-updates</option>:</term>
906 <indexterm><primary><option>-fwarn-incomplete-record-updates</option></primary></indexterm>
907 <indexterm><primary>incomplete record updates, warning</primary></indexterm>
908 <indexterm><primary>record updates, incomplete</primary></indexterm>
911 <function>f</function> below will fail when applied to
912 <literal>Bar</literal>, so the compiler will emit a warning about
913 this when <option>-fwarn-incomplete-record-updates</option> is
917 data Foo = Foo { x :: Int }
921 f foo = foo { x = 6 }
924 <para>This option isn't enabled be default because it can be
925 very noisy, and it often doesn't indicate a bug in the
932 <option>-fwarn-misc</option>:
933 <indexterm><primary><option>-fwarn-misc</option></primary></indexterm>
936 <para>Turns on warnings for various harmless but untidy
937 things. This currently includes: importing a type with
938 <literal>(..)</literal> when the export is abstract, and
939 listing duplicate class assertions in a qualified type.</para>
945 <option>-fwarn-missing-fields</option>:
946 <indexterm><primary><option>-fwarn-missing-fields</option></primary></indexterm>
947 <indexterm><primary>missing fields, warning</primary></indexterm>
948 <indexterm><primary>fields, missing</primary></indexterm>
952 <para>This option is on by default, and warns you whenever
953 the construction of a labelled field constructor isn't
954 complete, missing initializers for one or more fields. While
955 not an error (the missing fields are initialised with
956 bottoms), it is often an indication of a programmer error.</para>
961 <term><option>-fwarn-missing-methods</option>:</term>
963 <indexterm><primary><option>-fwarn-missing-methods</option></primary></indexterm>
964 <indexterm><primary>missing methods, warning</primary></indexterm>
965 <indexterm><primary>methods, missing</primary></indexterm>
967 <para>This option is on by default, and warns you whenever
968 an instance declaration is missing one or more methods, and
969 the corresponding class declaration has no default
970 declaration for them.</para>
971 <para>The warning is suppressed if the method name
972 begins with an underscore. Here's an example where this is useful:
975 _simpleFn :: a -> String
976 complexFn :: a -> a -> String
977 complexFn x y = ... _simpleFn ...
979 The idea is that: (a) users of the class will only call <literal>complexFn</literal>;
980 never <literal>_simpleFn</literal>; and (b)
981 instance declarations can define either <literal>complexFn</literal> or <literal>_simpleFn</literal>.
987 <term><option>-fwarn-missing-signatures</option>:</term>
989 <indexterm><primary><option>-fwarn-missing-signatures</option></primary></indexterm>
990 <indexterm><primary>type signatures, missing</primary></indexterm>
992 <para>If you would like GHC to check that every top-level
993 function/value has a type signature, use the
994 <option>-fwarn-missing-signatures</option> option. This
995 option is off by default.</para>
1000 <term><option>-fwarn-name-shadowing</option>:</term>
1002 <indexterm><primary><option>-fwarn-name-shadowing</option></primary></indexterm>
1003 <indexterm><primary>shadowing, warning</primary></indexterm>
1005 <para>This option causes a warning to be emitted whenever an
1006 inner-scope value has the same name as an outer-scope value,
1007 i.e. the inner value shadows the outer one. This can catch
1008 typographical errors that turn into hard-to-find bugs, e.g.,
1009 in the inadvertent cyclic definition <literal>let x = ... x
1010 ... in</literal>.</para>
1012 <para>Consequently, this option does
1013 <emphasis>will</emphasis> complain about cyclic recursive
1019 <term><option>-fwarn-orphans</option>:</term>
1021 <indexterm><primary><option>-fwarn-orphans</option></primary></indexterm>
1022 <indexterm><primary>orphan instances, warning</primary></indexterm>
1023 <indexterm><primary>orphan rules, warning</primary></indexterm>
1025 <para>This option causes a warning to be emitted whenever the
1026 module contains an "orphan" instance declaration or rewrite rule.
1027 An instance declartion is an orphan if it appears in a module in
1028 which neither the class nor the type being instanced are declared
1029 in the same module. A rule is an orphan if it is a rule for a
1030 function declared in another module. A module containing any
1031 orphans is called an orphan module.</para>
1032 <para>The trouble with orphans is that GHC must pro-actively read the interface
1033 files for all orphan modules, just in case their instances or rules
1034 play a role, whether or not the module's interface would otherwise
1035 be of any use. Other things being equal, avoid orphan modules.</para>
1041 <option>-fwarn-overlapping-patterns</option>:
1042 <indexterm><primary><option>-fwarn-overlapping-patterns</option></primary></indexterm>
1043 <indexterm><primary>overlapping patterns, warning</primary></indexterm>
1044 <indexterm><primary>patterns, overlapping</primary></indexterm>
1047 <para>By default, the compiler will warn you if a set of
1048 patterns are overlapping, i.e.,</para>
1051 f :: String -> Int
1057 <para>where the last pattern match in <function>f</function>
1058 won't ever be reached, as the second pattern overlaps
1059 it. More often than not, redundant patterns is a programmer
1060 mistake/error, so this option is enabled by default.</para>
1065 <term><option>-fwarn-simple-patterns</option>:</term>
1067 <indexterm><primary><option>-fwarn-simple-patterns</option></primary>
1069 <para>Causes the compiler to warn about lambda-bound
1070 patterns that can fail, eg. <literal>\(x:xs)->...</literal>.
1071 Normally, these aren't treated as incomplete patterns by
1072 <option>-fwarn-incomplete-patterns</option>.</para>
1073 <para>``Lambda-bound patterns'' includes all places where there is a single pattern,
1074 including list comprehensions and do-notation. In these cases, a pattern-match
1075 failure is quite legitimate, and triggers filtering (list comprehensions) or
1076 the monad <literal>fail</literal> operation (monads). For example:
1078 f :: [Maybe a] -> [a]
1079 f xs = [y | Just y <- xs]
1081 Switching on <option>-fwarn-simple-patterns</option> will elicit warnings about
1082 these probably-innocent cases, which is why the flag is off by default. </para>
1083 <para> The <literal>deriving( Read )</literal> mechanism produces monadic code with
1084 pattern matches, so you will also get misleading warnings about the compiler-generated
1085 code. (This is arguably a Bad Thing, but it's awkward to fix.)</para>
1091 <term><option>-fwarn-type-defaults</option>:</term>
1093 <indexterm><primary><option>-fwarn-type-defaults</option></primary></indexterm>
1094 <indexterm><primary>defaulting mechanism, warning</primary></indexterm>
1095 <para>Have the compiler warn/inform you where in your source
1096 the Haskell defaulting mechanism for numeric types kicks
1097 in. This is useful information when converting code from a
1098 context that assumed one default into one with another,
1099 e.g., the `default default' for Haskell 1.4 caused the
1100 otherwise unconstrained value <constant>1</constant> to be
1101 given the type <literal>Int</literal>, whereas Haskell 98
1102 defaults it to <literal>Integer</literal>. This may lead to
1103 differences in performance and behaviour, hence the
1104 usefulness of being non-silent about this.</para>
1106 <para>This warning is off by default.</para>
1111 <term><option>-fwarn-unused-binds</option>:</term>
1113 <indexterm><primary><option>-fwarn-unused-binds</option></primary></indexterm>
1114 <indexterm><primary>unused binds, warning</primary></indexterm>
1115 <indexterm><primary>binds, unused</primary></indexterm>
1116 <para>Report any function definitions (and local bindings)
1117 which are unused. For top-level functions, the warning is
1118 only given if the binding is not exported.</para>
1119 <para>A definition is regarded as "used" if (a) it is exported, or (b) it is
1120 mentioned in the right hand side of another definition that is used, or (c) the
1121 function it defines begins with an underscore. The last case provides a
1122 way to suppress unused-binding warnings selectively. </para>
1123 <para> Notice that a variable
1124 is reported as unused even if it appears in the right-hand side of another
1125 unused binding. </para>
1130 <term><option>-fwarn-unused-imports</option>:</term>
1132 <indexterm><primary><option>-fwarn-unused-imports</option></primary></indexterm>
1133 <indexterm><primary>unused imports, warning</primary></indexterm>
1134 <indexterm><primary>imports, unused</primary></indexterm>
1136 <para>Report any modules that are explicitly imported but
1137 never used. However, the form <literal>import M()</literal> is
1138 never reported as an unused import, because it is a useful idiom
1139 for importing instance declarations, which are anonymous in Haskell.</para>
1144 <term><option>-fwarn-unused-matches</option>:</term>
1146 <indexterm><primary><option>-fwarn-unused-matches</option></primary></indexterm>
1147 <indexterm><primary>unused matches, warning</primary></indexterm>
1148 <indexterm><primary>matches, unused</primary></indexterm>
1150 <para>Report all unused variables which arise from pattern
1151 matches, including patterns consisting of a single variable.
1152 For instance <literal>f x y = []</literal> would report
1153 <varname>x</varname> and <varname>y</varname> as unused. The
1154 warning is suppressed if the variable name begins with an underscore, thus:
1164 <para>If you're feeling really paranoid, the
1165 <option>-dcore-lint</option>
1166 option<indexterm><primary><option>-dcore-lint</option></primary></indexterm>
1167 is a good choice. It turns on heavyweight intra-pass
1168 sanity-checking within GHC. (It checks GHC's sanity, not
1175 <sect1 id="options-optimise">
1176 <title>Optimisation (code improvement)</title>
1178 <indexterm><primary>optimisation</primary></indexterm>
1179 <indexterm><primary>improvement, code</primary></indexterm>
1181 <para>The <option>-O*</option> options specify convenient
1182 “packages” of optimisation flags; the
1183 <option>-f*</option> options described later on specify
1184 <emphasis>individual</emphasis> optimisations to be turned on/off;
1185 the <option>-m*</option> options specify
1186 <emphasis>machine-specific</emphasis> optimisations to be turned
1189 <sect2 id="optimise-pkgs">
1190 <title><option>-O*</option>: convenient “packages” of optimisation flags.</title>
1192 <para>There are <emphasis>many</emphasis> options that affect
1193 the quality of code produced by GHC. Most people only have a
1194 general goal, something like “Compile quickly” or
1195 “Make my program run like greased lightning.” The
1196 following “packages” of optimisations (or lack
1197 thereof) should suffice.</para>
1199 <para>Note that higher optimisation levels cause more
1200 cross-module optimisation to be performed, which can have an
1201 impact on how much of your program needs to be recompiled when
1202 you change something. This is one reaosn to stick to
1203 no-optimisation when developing code.</para>
1209 No <option>-O*</option>-type option specified:
1210 <indexterm><primary>-O* not specified</primary></indexterm>
1213 <para>This is taken to mean: “Please compile
1214 quickly; I'm not over-bothered about compiled-code
1215 quality.” So, for example: <command>ghc -c
1216 Foo.hs</command></para>
1222 <option>-O0</option>:
1223 <indexterm><primary><option>-O0</option></primary></indexterm>
1226 <para>Means “turn off all optimisation”,
1227 reverting to the same settings as if no
1228 <option>-O</option> options had been specified. Saying
1229 <option>-O0</option> can be useful if
1230 eg. <command>make</command> has inserted a
1231 <option>-O</option> on the command line already.</para>
1237 <option>-O</option> or <option>-O1</option>:
1238 <indexterm><primary>-O option</primary></indexterm>
1239 <indexterm><primary>-O1 option</primary></indexterm>
1240 <indexterm><primary>optimise</primary><secondary>normally</secondary></indexterm>
1243 <para>Means: “Generate good-quality code without
1244 taking too long about it.” Thus, for example:
1245 <command>ghc -c -O Main.lhs</command></para>
1247 <para><option>-O</option> currently also implies
1248 <option>-fvia-C</option>. This may change in the
1255 <option>-O2</option>:
1256 <indexterm><primary>-O2 option</primary></indexterm>
1257 <indexterm><primary>optimise</primary><secondary>aggressively</secondary></indexterm>
1260 <para>Means: “Apply every non-dangerous
1261 optimisation, even if it means significantly longer
1262 compile times.”</para>
1264 <para>The avoided “dangerous” optimisations
1265 are those that can make runtime or space
1266 <emphasis>worse</emphasis> if you're unlucky. They are
1267 normally turned on or off individually.</para>
1269 <para>At the moment, <option>-O2</option> is
1270 <emphasis>unlikely</emphasis> to produce better code than
1271 <option>-O</option>.</para>
1277 <option>-Ofile <file></option>:
1278 <indexterm><primary>-Ofile <file> option</primary></indexterm>
1279 <indexterm><primary>optimising, customised</primary></indexterm>
1282 <para>(NOTE: not supported since GHC 4.x. Please ask if
1283 you're interested in this.)</para>
1285 <para>For those who need <emphasis>absolute</emphasis>
1286 control over <emphasis>exactly</emphasis> what options are
1287 used (e.g., compiler writers, sometimes :-), a list of
1288 options can be put in a file and then slurped in with
1289 <option>-Ofile</option>.</para>
1291 <para>In that file, comments are of the
1292 <literal>#</literal>-to-end-of-line variety; blank
1293 lines and most whitespace is ignored.</para>
1295 <para>Please ask if you are baffled and would like an
1296 example of <option>-Ofile</option>!</para>
1301 <para>We don't use a <option>-O*</option> flag for day-to-day
1302 work. We use <option>-O</option> to get respectable speed;
1303 e.g., when we want to measure something. When we want to go for
1304 broke, we tend to use <option>-O2 -fvia-C</option> (and we go for
1305 lots of coffee breaks).</para>
1307 <para>The easiest way to see what <option>-O</option> (etc.)
1308 “really mean” is to run with <option>-v</option>,
1309 then stand back in amazement.</para>
1312 <sect2 id="options-f">
1313 <title><option>-f*</option>: platform-independent flags</title>
1315 <indexterm><primary>-f* options (GHC)</primary></indexterm>
1316 <indexterm><primary>-fno-* options (GHC)</primary></indexterm>
1318 <para>These flags turn on and off individual optimisations.
1319 They are normally set via the <option>-O</option> options
1320 described above, and as such, you shouldn't need to set any of
1321 them explicitly (indeed, doing so could lead to unexpected
1322 results). However, there are one or two that may be of
1327 <term><option>-fexcess-precision</option>:</term>
1329 <indexterm><primary><option>-fexcess-precision</option></primary></indexterm>
1330 <para>When this option is given, intermediate floating
1331 point values can have a <emphasis>greater</emphasis>
1332 precision/range than the final type. Generally this is a
1333 good thing, but some programs may rely on the exact
1335 <literal>Float</literal>/<literal>Double</literal> values
1336 and should not use this option for their compilation.</para>
1341 <term><option>-fignore-asserts</option>:</term>
1343 <indexterm><primary><option>-fignore-asserts</option></primary></indexterm>
1344 <para>Causes GHC to ignore uses of the function
1345 <literal>Exception.assert</literal> in source code (in
1346 other words, rewriting <literal>Exception.assert p
1347 e</literal> to <literal>e</literal> (see <xref
1348 linkend="sec-assertions"/>). This flag is turned on by
1349 <option>-O</option>.
1356 <option>-fno-cse</option>
1357 <indexterm><primary><option>-fno-cse</option></primary></indexterm>
1360 <para>Turns off the common-sub-expression elimination optimisation.
1361 Can be useful if you have some <literal>unsafePerformIO</literal>
1362 expressions that you don't want commoned-up.</para>
1368 <option>-fno-strictness</option>
1369 <indexterm><primary><option>-fno-strictness</option></primary></indexterm>
1372 <para>Turns off the strictness analyser; sometimes it eats
1373 too many cycles.</para>
1379 <option>-fno-full-laziness</option>
1380 <indexterm><primary><option>-fno-full-laziness</option></primary></indexterm>
1383 <para>Turns off the full laziness optimisation (also known as
1384 let-floating). Full laziness increases sharing, which can lead
1385 to increased memory residency.</para>
1387 <para>NOTE: GHC doesn't implement complete full-laziness.
1388 When optimisation in on, and
1389 <option>-fno-full-laziness</option> is not given, some
1390 transformations that increase sharing are performed, such
1391 as extracting repeated computations from a loop. These
1392 are the same transformations that a fully lazy
1393 implementation would do, the difference is that GHC
1394 doesn't consistently apply full-laziness, so don't rely on
1401 <option>-fno-state-hack</option>
1402 <indexterm><primary><option>-fno-state-hack</option></primary></indexterm>
1405 <para>Turn off the "state hack" whereby any lambda with a
1406 <literal>State#</literal> token as argument is considered to be
1407 single-entry, hence it is considered OK to inline things inside
1408 it. This can improve performance of IO and ST monad code, but it
1409 runs the risk of reducing sharing.</para>
1415 <option>-funbox-strict-fields</option>:
1416 <indexterm><primary><option>-funbox-strict-fields</option></primary></indexterm>
1417 <indexterm><primary>strict constructor fields</primary></indexterm>
1418 <indexterm><primary>constructor fields, strict</primary></indexterm>
1421 <para>This option causes all constructor fields which are
1422 marked strict (i.e. “!”) to be unboxed or
1423 unpacked if possible. It is equivalent to adding an
1424 <literal>UNPACK</literal> pragma to every strict
1425 constructor field (see <xref
1426 linkend="unpack-pragma"/>).</para>
1428 <para>This option is a bit of a sledgehammer: it might
1429 sometimes make things worse. Selectively unboxing fields
1430 by using <literal>UNPACK</literal> pragmas might be
1437 <option>-funfolding-update-in-place<n></option>
1438 <indexterm><primary><option>-funfolding-update-in-place</option></primary></indexterm>
1441 <para>Switches on an experimental "optimisation".
1442 Switching it on makes the compiler a little keener to
1443 inline a function that returns a constructor, if the
1444 context is that of a thunk.
1448 If we inlined plusInt we might get an opportunity to use
1449 update-in-place for the thunk 'x'.</para>
1455 <option>-funfolding-creation-threshold<n></option>:
1456 <indexterm><primary><option>-funfolding-creation-threshold</option></primary></indexterm>
1457 <indexterm><primary>inlining, controlling</primary></indexterm>
1458 <indexterm><primary>unfolding, controlling</primary></indexterm>
1461 <para>(Default: 45) Governs the maximum size that GHC will
1462 allow a function unfolding to be. (An unfolding has a
1463 “size” that reflects the cost in terms of
1464 “code bloat” of expanding that unfolding at
1465 at a call site. A bigger function would be assigned a
1466 bigger cost.) </para>
1468 <para> Consequences: (a) nothing larger than this will be
1469 inlined (unless it has an INLINE pragma); (b) nothing
1470 larger than this will be spewed into an interface
1474 <para> Increasing this figure is more likely to result in longer
1475 compile times than faster code. The next option is more
1481 <term><option>-funfolding-use-threshold<n></option>:</term>
1483 <indexterm><primary><option>-funfolding-use-threshold</option></primary></indexterm>
1484 <indexterm><primary>inlining, controlling</primary></indexterm>
1485 <indexterm><primary>unfolding, controlling</primary></indexterm>
1487 <para>(Default: 8) This is the magic cut-off figure for
1488 unfolding: below this size, a function definition will be
1489 unfolded at the call-site, any bigger and it won't. The
1490 size computed for a function depends on two things: the
1491 actual size of the expression minus any discounts that
1492 apply (see <option>-funfolding-con-discount</option>).</para>
1503 <sect1 id="sec-using-concurrent">
1504 <title>Using Concurrent Haskell</title>
1505 <indexterm><primary>Concurrent Haskell</primary><secondary>using</secondary></indexterm>
1507 <para>GHC supports Concurrent Haskell by default, without requiring a
1508 special option or libraries compiled in a certain way. To get access to
1509 the support libraries for Concurrent Haskell, just import
1511 url="../libraries/base/Control-Concurrent.html"><literal>Control.Concurrent</literal></ulink>. More information on Concurrent Haskell is provided in the documentation for that module.</para>
1513 <para>The following RTS option(s) affect the behaviour of Concurrent
1514 Haskell programs:<indexterm><primary>RTS options, concurrent</primary></indexterm></para>
1518 <term><option>-C<replaceable>s</replaceable></option></term>
1520 <para><indexterm><primary><option>-C<replaceable>s</replaceable></option></primary><secondary>RTS option</secondary></indexterm>
1521 Sets the context switch interval to <replaceable>s</replaceable>
1522 seconds. A context switch will occur at the next heap block
1523 allocation after the timer expires (a heap block allocation occurs
1524 every 4k of allocation). With <option>-C0</option> or
1525 <option>-C</option>, context switches will occur as often as
1526 possible (at every heap block allocation). By default, context
1527 switches occur every 20ms. Note that GHC's internal timer ticks
1528 every 20ms, and the context switch timer is always a multiple of
1529 this timer, so 20ms is the maximum granularity available for timed
1530 context switches.</para>
1536 <sect1 id="sec-using-parallel">
1537 <title>Using parallel Haskell</title>
1540 <indexterm><primary>Parallel Haskell</primary><secondary>using</secondary></indexterm>
1541 [NOTE: GHC does not support Parallel Haskell by default, you need to
1542 obtain a special version of GHC from the <ulink
1543 url="http://www.cee.hw.ac.uk/~dsg/gph/">GPH</ulink> site. Also,
1544 you won't be able to execute parallel Haskell programs unless PVM3
1545 (parallel Virtual Machine, version 3) is installed at your site.]
1549 To compile a Haskell program for parallel execution under PVM, use the
1550 <option>-parallel</option> option,<indexterm><primary>-parallel
1551 option</primary></indexterm> both when compiling <emphasis>and
1552 linking</emphasis>. You will probably want to <literal>import
1553 Control.Parallel</literal> into your Haskell modules.
1557 To run your parallel program, once PVM is going, just invoke it
1558 “as normal”. The main extra RTS option is
1559 <option>-qp<n></option>, to say how many PVM
1560 “processors” your program to run on. (For more details of
1561 all relevant RTS options, please see <xref
1562 linkend="parallel-rts-opts"/>.)
1566 In truth, running parallel Haskell programs and getting information
1567 out of them (e.g., parallelism profiles) is a battle with the vagaries of
1568 PVM, detailed in the following sections.
1571 <sect2 id="pvm-dummies">
1572 <title>Dummy's guide to using PVM</title>
1575 <indexterm><primary>PVM, how to use</primary></indexterm>
1576 <indexterm><primary>parallel Haskell—PVM use</primary></indexterm>
1577 Before you can run a parallel program under PVM, you must set the
1578 required environment variables (PVM's idea, not ours); something like,
1579 probably in your <filename>.cshrc</filename> or equivalent:
1582 setenv PVM_ROOT /wherever/you/put/it
1583 setenv PVM_ARCH `$PVM_ROOT/lib/pvmgetarch`
1584 setenv PVM_DPATH $PVM_ROOT/lib/pvmd
1590 Creating and/or controlling your “parallel machine” is a purely-PVM
1591 business; nothing specific to parallel Haskell. The following paragraphs
1592 describe how to configure your parallel machine interactively.
1596 If you use parallel Haskell regularly on the same machine configuration it
1597 is a good idea to maintain a file with all machine names and to make the
1598 environment variable PVM_HOST_FILE point to this file. Then you can avoid
1599 the interactive operations described below by just saying
1607 You use the <command>pvm</command><indexterm><primary>pvm command</primary></indexterm> command to start PVM on your
1608 machine. You can then do various things to control/monitor your
1609 “parallel machine;” the most useful being:
1615 <colspec align="left"/>
1619 <entry><keycombo><keycap>Control</keycap><keycap>D</keycap></keycombo></entry>
1620 <entry>exit <command>pvm</command>, leaving it running</entry>
1624 <entry><command>halt</command></entry>
1625 <entry>kill off this “parallel machine” & exit</entry>
1629 <entry><command>add <host></command></entry>
1630 <entry>add <command><host></command> as a processor</entry>
1634 <entry><command>delete <host></command></entry>
1635 <entry>delete <command><host></command></entry>
1639 <entry><command>reset</command></entry>
1640 <entry>kill what's going, but leave PVM up</entry>
1644 <entry><command>conf</command></entry>
1645 <entry>list the current configuration</entry>
1649 <entry><command>ps</command></entry>
1650 <entry>report processes' status</entry>
1654 <entry><command>pstat <pid></command></entry>
1655 <entry>status of a particular process</entry>
1664 The PVM documentation can tell you much, much more about <command>pvm</command>!
1669 <sect2 id="par-profiles">
1670 <title>parallelism profiles</title>
1673 <indexterm><primary>parallelism profiles</primary></indexterm>
1674 <indexterm><primary>profiles, parallelism</primary></indexterm>
1675 <indexterm><primary>visualisation tools</primary></indexterm>
1679 With parallel Haskell programs, we usually don't care about the
1680 results—only with “how parallel” it was! We want pretty pictures.
1684 parallelism profiles (à la <command>hbcpp</command>) can be generated with the
1685 <option>-qP</option><indexterm><primary>-qP RTS option</primary></indexterm> RTS option. The
1686 per-processor profiling info is dumped into files named
1687 <filename><full-path><program>.gr</filename>. These are then munged into a PostScript picture,
1688 which you can then display. For example, to run your program
1689 <filename>a.out</filename> on 8 processors, then view the parallelism profile, do:
1695 <prompt>$</prompt> ./a.out +RTS -qP -qp8
1696 <prompt>$</prompt> grs2gr *.???.gr > temp.gr # combine the 8 .gr files into one
1697 <prompt>$</prompt> gr2ps -O temp.gr # cvt to .ps; output in temp.ps
1698 <prompt>$</prompt> ghostview -seascape temp.ps # look at it!
1704 The scripts for processing the parallelism profiles are distributed
1705 in <filename>ghc/utils/parallel/</filename>.
1711 <title>Other useful info about running parallel programs</title>
1714 The “garbage-collection statistics” RTS options can be useful for
1715 seeing what parallel programs are doing. If you do either
1716 <option>+RTS -Sstderr</option><indexterm><primary>-Sstderr RTS option</primary></indexterm> or <option>+RTS -sstderr</option>, then
1717 you'll get mutator, garbage-collection, etc., times on standard
1718 error. The standard error of all PE's other than the `main thread'
1719 appears in <filename>/tmp/pvml.nnn</filename>, courtesy of PVM.
1723 Whether doing <option>+RTS -Sstderr</option> or not, a handy way to watch
1724 what's happening overall is: <command>tail -f /tmp/pvml.nnn</command>.
1729 <sect2 id="parallel-rts-opts">
1730 <title>RTS options for Parallel Haskell
1734 <indexterm><primary>RTS options, parallel</primary></indexterm>
1735 <indexterm><primary>parallel Haskell—RTS options</primary></indexterm>
1739 Besides the usual runtime system (RTS) options
1740 (<xref linkend="runtime-control"/>), there are a few options particularly
1741 for parallel execution.
1748 <term><option>-qp<N></option>:</term>
1751 <indexterm><primary>-qp<N> RTS option</primary></indexterm>
1752 (paraLLEL ONLY) Use <literal><N></literal> PVM processors to run this program;
1758 <term><option>-C[<s>]</option>:</term>
1761 <indexterm><primary>-C<s> RTS option</primary></indexterm> Sets
1762 the context switch interval to <literal><s></literal> seconds.
1763 A context switch will occur at the next heap block allocation after
1764 the timer expires (a heap block allocation occurs every 4k of
1765 allocation). With <option>-C0</option> or <option>-C</option>,
1766 context switches will occur as often as possible (at every heap block
1767 allocation). By default, context switches occur every 20ms. Note that GHC's internal timer ticks every 20ms, and
1768 the context switch timer is always a multiple of this timer, so 20ms
1769 is the maximum granularity available for timed context switches.
1774 <term><option>-q[v]</option>:</term>
1777 <indexterm><primary>-q RTS option</primary></indexterm>
1778 (paraLLEL ONLY) Produce a quasi-parallel profile of thread activity,
1779 in the file <filename><program>.qp</filename>. In the style of <command>hbcpp</command>, this profile
1780 records the movement of threads between the green (runnable) and red
1781 (blocked) queues. If you specify the verbose suboption (<option>-qv</option>), the
1782 green queue is split into green (for the currently running thread
1783 only) and amber (for other runnable threads). We do not recommend
1784 that you use the verbose suboption if you are planning to use the
1785 <command>hbcpp</command> profiling tools or if you are context switching at every heap
1786 check (with <option>-C</option>).
1792 <term><option>-qt<num></option>:</term>
1795 <indexterm><primary>-qt<num> RTS option</primary></indexterm>
1796 (paraLLEL ONLY) Limit the thread pool size, i.e. the number of
1797 threads per processor to <literal><num></literal>. The default is
1798 32. Each thread requires slightly over 1K <emphasis>words</emphasis> in
1799 the heap for thread state and stack objects. (For 32-bit machines, this
1800 translates to 4K bytes, and for 64-bit machines, 8K bytes.)
1806 <term><option>-d</option>:</term>
1809 <indexterm><primary>-d RTS option (parallel)</primary></indexterm>
1810 (paraLLEL ONLY) Turn on debugging. It pops up one xterm (or GDB, or
1811 something…) per PVM processor. We use the standard <command>debugger</command>
1812 script that comes with PVM3, but we sometimes meddle with the
1813 <command>debugger2</command> script. We include ours in the GHC distribution,
1814 in <filename>ghc/utils/pvm/</filename>.
1820 <term><option>-qe<num></option>:</term>
1823 <indexterm><primary>-qe<num> RTS option
1824 (parallel)</primary></indexterm> (paraLLEL ONLY) Limit the spark pool size
1825 i.e. the number of pending sparks per processor to
1826 <literal><num></literal>. The default is 100. A larger number may be
1827 appropriate if your program generates large amounts of parallelism
1833 <term><option>-qQ<num></option>:</term>
1836 <indexterm><primary>-qQ<num> RTS option (parallel)</primary></indexterm>
1837 (paraLLEL ONLY) Set the size of packets transmitted between processors
1838 to <literal><num></literal>. The default is 1024 words. A larger number may be
1839 appropriate if your machine has a high communication cost relative to
1845 <term><option>-qh<num></option>:</term>
1848 <indexterm><primary>-qh<num> RTS option (parallel)</primary></indexterm>
1849 (paraLLEL ONLY) Select a packing scheme. Set the number of non-root thunks to pack in one packet to
1850 <num>-1 (0 means infinity). By default GUM uses full-subgraph
1851 packing, i.e. the entire subgraph with the requested closure as root is
1852 transmitted (provided it fits into one packet). Choosing a smaller value
1853 reduces the amount of pre-fetching of work done in GUM. This can be
1854 advantageous for improving data locality but it can also worsen the balance
1855 of the load in the system.
1860 <term><option>-qg<num></option>:</term>
1863 <indexterm><primary>-qg<num> RTS option
1864 (parallel)</primary></indexterm> (paraLLEL ONLY) Select a globalisation
1865 scheme. This option affects the
1866 generation of global addresses when transferring data. Global addresses are
1867 globally unique identifiers required to maintain sharing in the distributed
1868 graph structure. Currently this is a binary option. With <num>=0 full globalisation is used
1869 (default). This means a global address is generated for every closure that
1870 is transmitted. With <num>=1 a thunk-only globalisation scheme is
1871 used, which generated global address only for thunks. The latter case may
1872 lose sharing of data but has a reduced overhead in packing graph structures
1873 and maintaining internal tables of global addresses.
1884 <sect1 id="options-platform">
1885 <title>Platform-specific Flags</title>
1887 <indexterm><primary>-m* options</primary></indexterm>
1888 <indexterm><primary>platform-specific options</primary></indexterm>
1889 <indexterm><primary>machine-specific options</primary></indexterm>
1891 <para>Some flags only make sense for particular target
1897 <term><option>-mv8</option>:</term>
1899 <para>(SPARC machines)<indexterm><primary>-mv8 option (SPARC
1900 only)</primary></indexterm> Means to pass the like-named
1901 option to GCC; it says to use the Version 8 SPARC
1902 instructions, notably integer multiply and divide. The
1903 similar <option>-m*</option> GCC options for SPARC also
1904 work, actually.</para>
1909 <term><option>-monly-[32]-regs</option>:</term>
1911 <para>(iX86 machines)<indexterm><primary>-monly-N-regs
1912 option (iX86 only)</primary></indexterm> GHC tries to
1913 “steal” four registers from GCC, for performance
1914 reasons; it almost always works. However, when GCC is
1915 compiling some modules with four stolen registers, it will
1916 crash, probably saying:
1919 Foo.hc:533: fixed or forbidden register was spilled.
1920 This may be due to a compiler bug or to impossible asm
1921 statements or clauses.
1924 Just give some registers back with
1925 <option>-monly-N-regs</option>. Try `3' first, then `2'.
1926 If `2' doesn't work, please report the bug to us.</para>
1935 <sect1 id="ext-core">
1936 <title>Generating and compiling External Core Files</title>
1938 <indexterm><primary>intermediate code generation</primary></indexterm>
1940 <para>GHC can dump its optimized intermediate code (said to be in “Core” format)
1941 to a file as a side-effect of compilation. Core files, which are given the suffix
1942 <filename>.hcr</filename>, can be read and processed by non-GHC back-end
1943 tools. The Core format is formally described in <ulink url="http://www.haskell.org/ghc/docs/papers/core.ps.gz">
1944 <citetitle>An External Representation for the GHC Core Language</citetitle></ulink>,
1945 and sample tools (in Haskell)
1946 for manipulating Core files are available in the GHC source distribution
1947 directory <literal>/fptools/ghc/utils/ext-core</literal>.
1948 Note that the format of <literal>.hcr</literal>
1949 files is <emphasis>different</emphasis> (though similar) to the Core output format generated
1950 for debugging purposes (<xref linkend="options-debugging"/>).</para>
1952 <para>The Core format natively supports notes which you can add to
1953 your source code using the <literal>CORE</literal> pragma (see <xref
1954 linkend="pragmas"/>).</para>
1960 <option>-fext-core</option>
1961 <indexterm><primary><option>-fext-core</option></primary></indexterm>
1964 <para>Generate <literal>.hcr</literal> files.</para>
1970 <para>GHC can also read in External Core files as source; just give the <literal>.hcr</literal> file on
1971 the command line, instead of the <literal>.hs</literal> or <literal>.lhs</literal> Haskell source.
1972 A current infelicity is that you need to give the <literal>-fglasgow-exts</literal> flag too, because
1973 ordinary Haskell 98, when translated to External Core, uses things like rank-2 types.</para>
1982 ;;; Local Variables: ***
1984 ;;; sgml-parent-document: ("users_guide.xml" "book" "chapter") ***