1 <chapter id="using-ghc">
2 <title>Using GHC</title>
4 <indexterm><primary>GHC, using</primary></indexterm>
5 <indexterm><primary>using GHC</primary></indexterm>
7 <para>GHC can work in one of three “modes”:</para>
11 <term><cmdsynopsis><command>ghc</command>
12 <arg choice=plain>--interactive</arg>
14 <indexterm><primary>interactive mode</primary>
16 <indexterm><primary>ghci</primary>
19 <para>Interactive mode, which is also available as
20 <command>ghci</command>. Interactive mode is described in
21 more detail in <xref linkend="ghci">.</para>
26 <term><cmdsynopsis><command>ghc</command>
27 <arg choice=plain>--make</arg>
29 <indexterm><primary>make mode</primary>
31 <indexterm><primary><option>--make</option></primary>
34 <para>In this mode, GHC will build a multi-module Haskell
35 program automatically, figuring out dependencies for itself.
36 If you have a straightforward Haskell program, this is
37 likely to be much easier, and faster, than using
38 <command>make</command>.</para>
44 <command>ghc</command>
52 <indexterm><primary><option>--make</option></primary>
55 <para>This is the traditional batch-compiler mode, in which
56 GHC can compile source files one at a time, or link objects
57 together into an executable.</para>
63 <title>Options overview</title>
65 <para>GHC's behaviour is controlled by
66 <firstterm>options</firstterm>, which for historical reasons are
67 also sometimes referred to as command-line flags or arguments.
68 Options can be specified in three ways:</para>
71 <title>Command-line arguments</title>
73 <indexterm><primary>structure, command-line</primary></indexterm>
74 <indexterm><primary>command-line</primary><secondary>arguments</secondary></indexterm>
75 <indexterm><primary>arguments</primary><secondary>command-line</secondary></indexterm>
77 <para>An invocation of GHC takes the following form:</para>
83 <para>Command-line arguments are either options or file names.</para>
85 <para>Command-line options begin with <literal>-</literal>.
86 They may <emphasis>not</emphasis> be grouped:
87 <option>-vO</option> is different from <option>-v -O</option>.
88 Options need not precede filenames: e.g., <literal>ghc *.o -o
89 foo</literal>. All options are processed and then applied to
90 all files; you cannot, for example, invoke <literal>ghc -c -O1
91 Foo.hs -O2 Bar.hs</literal> to apply different optimisation
92 levels to the files <filename>Foo.hs</filename> and
93 <filename>Bar.hs</filename>.</para>
96 <Sect2 id="source-file-options">
97 <title>Command line options in source files</title>
99 <indexterm><primary>source-file options</primary></indexterm>
101 <para>Sometimes it is useful to make the connection between a
102 source file and the command-line options it requires quite
103 tight. For instance, if a Haskell source file uses GHC
104 extensions, it will always need to be compiled with the
105 <option>-fglasgow-exts</option> option. Rather than maintaining
106 the list of per-file options in a <filename>Makefile</filename>,
107 it is possible to do this directly in the source file using the
108 <literal>OPTIONS</literal> pragma <indexterm><primary>OPTIONS
109 pragma</primary></indexterm>:</para>
112 {-# OPTIONS -fglasgow-exts #-}
117 <para><literal>OPTIONS</literal> pragmas are only looked for at
118 the top of your source files, upto the first
119 (non-literate,non-empty) line not containing
120 <literal>OPTIONS</literal>. Multiple <literal>OPTIONS</literal>
121 pragmas are recognised. Note that your command shell does not
122 get to the source file options, they are just included literally
123 in the array of command-line arguments the compiler driver
124 maintains internally, so you'll be desperately disappointed if
125 you try to glob etc. inside <literal>OPTIONS</literal>.</para>
127 <para>NOTE: the contents of OPTIONS are prepended to the
128 command-line options, so you <emphasis>do</emphasis> have the
129 ability to override OPTIONS settings via the command
132 <para>It is not recommended to move all the contents of your
133 Makefiles into your source files, but in some circumstances, the
134 <literal>OPTIONS</literal> pragma is the Right Thing. (If you
135 use <option>-keep-hc-file-too</option> and have OPTION flags in
136 your module, the OPTIONS will get put into the generated .hc
141 <title>Setting options in GHCi</title>
143 <para>Options may also be modified from within GHCi, using the
144 <literal>:set</literal> command. See <xref linkend="ghci-set">
145 for more details.</para>
149 <sect1 id="static-dynamic-flags">
150 <title>Static vs. Dynamic options</title>
151 <indexterm><primary>static</primary><secondary>options</secondary>
153 <indexterm><primary>dynamic</primary><secondary>options</secondary>
156 <para>Each of GHC's command line options is classified as either
157 <firstterm>static</firstterm> or <firstterm>dynamic</firstterm>.
158 A static flag may only be specified on the command line, whereas a
159 dynamic flag may also be given in an <literal>OPTIONS</literal>
160 pragma in a source file or set from the GHCi command-line with
161 <literal>:set</literal>.</para>
163 <para>As a rule of thumb, all the language options are dynamic, as
164 are the warning options and the debugging options. The rest are
165 static, with the notable exceptions of <option>-v</option>,
166 <option>-cpp</option>, <option>-fasm</option>,
167 <option>-fvia-C</option>, and <option>-#include</option>.
169 The flag reference tables (<xref linkend="flag-reference">) lists
170 the status of each flag.</para>
173 <sect1 id="file-suffixes">
174 <title>Meaningful file suffixes</title>
176 <indexterm><primary>suffixes, file</primary></indexterm>
177 <indexterm><primary>file suffixes for GHC</primary></indexterm>
179 <para>File names with “meaningful” suffixes (e.g.,
180 <filename>.lhs</filename> or <filename>.o</filename>) cause the
181 “right thing” to happen to those files.</para>
186 <term><filename>.lhs</filename></term>
187 <indexterm><primary><literal>lhs</literal> suffix</primary></indexterm>
189 <para>A “literate Haskell” module.</para>
194 <term><filename>.hs</filename></term>
196 <para>A not-so-literate Haskell module.</para>
201 <term><filename>.hi</filename></term>
203 <para>A Haskell interface file, probably
204 compiler-generated.</para>
209 <term><filename>.hc</filename></term>
211 <para>Intermediate C file produced by the Haskell
217 <term><filename>.c</filename></term>
219 <para>A C file not produced by the Haskell
225 <term><filename>.s</filename></term>
227 <para>An assembly-language source file, usually produced by
233 <term><filename>.o</filename></term>
235 <para>An object file, produced by an assembler.</para>
240 <para>Files with other suffixes (or without suffixes) are passed
241 straight to the linker.</para>
245 <sect1 id="options-help">
246 <title>Help and verbosity options</title>
248 <IndexTerm><Primary>help options</Primary></IndexTerm>
249 <IndexTerm><Primary>verbosity options</Primary></IndexTerm>
253 <term><literal>-help</literal></term>
254 <term><literal>-?</literal></term>
255 <indexterm><primary><literal>-?</literal></primary></indexterm>
256 <indexterm><primary><literal>-help</literal></primary></indexterm>
258 <para>Cause GHC to spew a long usage message to standard
259 output and then exit.</para>
264 <term><literal>-v</literal></term>
265 <indexterm><primary><literal>-v</literal></primary></indexterm>
267 <para>The <option>-v</option> option makes GHC
268 <emphasis>verbose</emphasis>: it reports its version number
269 and shows (on stderr) exactly how it invokes each phase of
270 the compilation system. Moreover, it passes the
271 <option>-v</option> flag to most phases; each reports its
272 version number (and possibly some other information).</para>
274 <para>Please, oh please, use the <option>-v</option> option
275 when reporting bugs! Knowing that you ran the right bits in
276 the right order is always the first thing we want to
282 <term><literal>-v</literal><replaceable>n</replaceable></term>
283 <indexterm><primary><option>-v</option></primary></indexterm>
285 <para>To provide more control over the compiler's verbosity,
286 the <option>-v</option> flag takes an optional numeric
287 argument. Specifying <option>-v</option> on its own is
288 equivalent to <option>-v3</option>, and the other levels
289 have the following meanings:</para>
293 <term><literal>-v0</literal></term>
295 <para>Disable all non-essential messages (this is the
301 <term><literal>-v1</literal></term>
303 <para>Minimal verbosity: print one line per
304 compilation (this is the default when
305 <option>--make</option> or
306 <option>--interactive</option> is on).</para>
311 <term><literal>-v2</literal></term>
313 <para>Print the name of each compilation phase as it
314 is executed. (equivalent to
315 <option>-dshow-passes</option>).</para>
320 <term><literal>-v3</literal></term>
322 <para>The same as <option>-v2</option>, except that in
323 addition the full command line (if appropriate) for
324 each compilation phase is also printed.</para>
329 <term><literal>-v4</literal></term>
331 <para>The same as <option>-v3</option> except that the
332 intermediate program representation after each
333 compilation phase is also printed (excluding
334 preprocessed and C/assembly files).</para>
342 <term><literal>--version</literal></term>
343 <indexterm><primary><literal>--version</literal></primary></indexterm>
345 <para>Print a one-line string including GHC's version number.</para>
350 <term><literal>--numeric-version</literal></term>
351 <indexterm><primary><literal>--numeric-version</literal></primary></indexterm>
353 <para>Print GHC's numeric version number only.</para>
359 <sect1 id="make-mode">
360 <title>Using <command>ghc</command> <option>--make</option></title>
362 <indexterm><primary><option>--make</option></primary>
364 <indexterm><primary>separate compilation</primary>
367 <para>When given the <option>--make</option> option, GHC will
368 build a multi-module Haskell program by following dependencies
369 from a single root module (usually <literal>Main</literal>). For
370 example, if your <literal>Main</literal> module is in a file
371 called <filename>Main.hs</filename>, you could compile and link
372 the program like this:</para>
378 <para>The command line must contain one source file or module
379 name; GHC will figure out all the modules in the program by
380 following the imports from this initial module. It will then
381 attempt to compile each module which is out of date, and finally
382 if the top module is <literal>Main</literal>, the program
383 will also be linked into an executable.</para>
385 <para>The main advantages to using <literal>ghc --make</literal>
386 over traditional <literal>Makefile</literal>s are:</para>
390 <para>GHC doesn't have to be restarted for each compilation,
391 which means it can cache information between compilations.
392 Compiling a muli-module program with <literal>ghc
393 --make</literal> can be up to twice as fast as running
394 <literal>ghc</literal> individually on each source
398 <para>You don't have to write a
399 <literal>Makefile</literal>.</para>
401 <indexterm><primary><literal>Makefile</literal>s</primary><secondary>avoiding</secondary>
404 <para>GHC re-calculates the dependencies each time it is
405 invoked, so the dependencies never get out of sync with the
410 <para>Any of the command-line options described in the rest of
411 this chapter can be used with <option>--make</option>, but note
412 that any options you give on the command line will apply to all
413 the source files compiled, so if you want any options to apply to
414 a single source file only, you'll need to use an
415 <literal>OPTIONS</literal> pragma (see <xref
416 linkend="source-file-options">).</para>
418 <para>If the program needs to be linked with additional objects
419 (say, some auxilliary C code), these can be specified on the
420 command line as usual.</para>
423 <Sect1 id="options-order">
424 <title>GHC without <option>--make</option></title>
426 <para>Without <option>--make</option>, GHC will compile one or
427 more source files given on the command line.</para>
429 <para>The first phase to run is determined by each input-file
430 suffix, and the last phase is determined by a flag. If no
431 relevant flag is present, then go all the way through linking.
432 This table summarises:</para>
436 <colspec align="left">
437 <colspec align="left">
438 <colspec align="left">
439 <colspec align="left">
443 <entry>Phase of the compilation system</entry>
444 <entry>Suffix saying “start here”</entry>
445 <entry>Flag saying “stop after”</entry>
446 <entry>(suffix of) output file</entry>
451 <entry>literate pre-processor</entry>
452 <entry><literal>.lhs</literal></entry>
454 <entry><literal>.hs</literal></entry>
458 <entry>C pre-processor (opt.)
460 <entry><literal>.hs</literal> (with
461 <option>-cpp</option>)</entry>
462 <entry><option>-E</option></entry>
463 <entry><literal>.hspp</literal></entry>
467 <entry>Haskell compiler</entry>
468 <entry><literal>.hs</literal></entry>
469 <entry><option>-C</option>, <option>-S</option></entry>
470 <entry><literal>.hc</literal>, <literal>.s</literal></entry>
474 <entry>C compiler (opt.)</entry>
475 <entry><literal>.hc</literal> or <literal>.c</literal></entry>
476 <entry><option>-S</option></entry>
477 <entry><literal>.s</literal></entry>
481 <entry>assembler</entry>
482 <entry><literal>.s</literal></entry>
483 <entry><option>-c</option></entry>
484 <entry><literal>.o</literal></entry>
488 <entry>linker</entry>
489 <entry><replaceable>other</replaceable></entry>
491 <entry><filename>a.out</filename></entry>
497 <indexterm><primary><option>-C</option></primary></indexterm>
498 <indexterm><primary><option>-E</option></primary></indexterm>
499 <indexterm><primary><option>-S</option></primary></indexterm>
500 <indexterm><primary><option>-c</option></primary></indexterm>
502 <para>Thus, a common invocation would be: <literal>ghc -c
503 Foo.hs</literal></para>
505 <para>Note: What the Haskell compiler proper produces depends on
506 whether a native-code generator<indexterm><primary>native-code
507 generator</primary></indexterm> is used (producing assembly
508 language) or not (producing C). See <xref
509 linkend="options-codegen"> for more details.</para>
511 <para>Note: C pre-processing is optional, the
512 <option>-ccp</option><indexterm><primary><option>-cpp</option></primary>
513 </indexterm>flag turns it on. See <xref
514 linkend="c-pre-processor"> for more details.</para>
516 <para>Note: The option <option>-E</option><IndexTerm><Primary>-E
517 option</Primary></IndexTerm> runs just the pre-processing passes
518 of the compiler, dumping the result in a file. Note that this
519 differs from the previous behaviour of dumping the file to
520 standard output.</para>
523 <sect1 id="options-output">
524 <title>Re-directing the compilation output(s)</title>
526 <indexterm><primary>output-directing options</primary></indexterm>
527 <indexterm><primary>redirecting compilation output</primary></indexterm>
532 <term><literal>-o</literal></term>
533 <indexterm><primary><literal>-o</literal></primary></indexterm>
535 <para>GHC's compiled output normally goes into a
536 <filename>.hc</filename>, <filename>.o</filename>, etc.,
537 file, depending on the last-run compilation phase. The
538 option <option>-o foo</option><IndexTerm><Primary>-o
539 option</Primary></IndexTerm> re-directs the output of that
540 last-run phase to file <filename>foo</filename>.</para>
542 <para>Note: this “feature” can be
543 counterintuitive: <command>ghc -C -o foo.o foo.hs</command>
544 will put the intermediate C code in the file
545 <filename>foo.o</filename>, name notwithstanding!</para>
550 <term><literal>-odir</literal></term>
551 <indexterm><primary><literal>-odir</literal></primary></indexterm>
553 <para>The <option>-o</option> option isn't of much use if
554 you have <emphasis>several</emphasis> input files…
555 Non-interface output files are normally put in the same
556 directory as their corresponding input file came from. You
557 may specify that they be put in another directory using the
558 <option>-odir <dir></option><IndexTerm><Primary>-odir
559 <dir> option</Primary></IndexTerm> (the “Oh,
560 dear” option). For example:</para>
563 % ghc -c parse/Foo.hs parse/Bar.hs gurgle/Bumble.hs -odir `arch`
566 <para>The output files, <filename>Foo.o</filename>,
567 <filename>Bar.o</filename>, and
568 <filename>Bumble.o</filename> would be put into a
569 subdirectory named after the architecture of the executing
570 machine (<filename>sun4</filename>,
571 <filename>mips</filename>, etc). The directory must already
572 exist; it won't be created.</para>
574 <para>Note that the <option>-odir</option> option does
575 <emphasis>not</emphasis> affect where the interface files
576 are put. In the above example, they would still be put in
577 <filename>parse/Foo.hi</filename>,
578 <filename>parse/Bar.hi</filename>, and
579 <filename>gurgle/Bumble.hi</filename>.</para>
584 <term><literal>-ohi</literal></term>
585 <indexterm><primary><literal>-ohi</literal></primary></indexterm>
592 <term><literal>-osuf</literal></term>
593 <term><literal>-hisuf</literal></term>
594 <indexterm><primary><literal>-osuf</literal></primary></indexterm>
595 <indexterm><primary><literal>-hisuf</literal></primary></indexterm>
597 <para>EXOTICA: The <option>-osuf
598 <suffix></option><IndexTerm><Primary>-osuf
599 <suffix> option</Primary></IndexTerm> will change the
600 <filename>.o</filename> file suffix for object files to
601 whatever you specify. (We use this in compiling the
604 <para>Similarly, the <option>-hisuf
605 <suffix></option><IndexTerm><Primary>-hisuf
606 <suffix> option</Primary></IndexTerm> will change the
607 <filename>.hi</filename> file suffix for non-system
608 interface files (see <XRef LinkEnd="hi-options">).</para>
610 <para>The <option>-hisuf</option>/<option>-osuf</option>
611 game is useful if you want to compile a program with both
612 GHC and HBC (say) in the same directory. Let HBC use the
613 standard <filename>.hi</filename>/<filename>.o</filename>
614 suffixes; add <option>-hisuf g_hi -osuf
615 g_o</option> to your <command>make</command> rule for
616 GHC compiling…</para>
621 <sect2 id="keeping-intermediates">
622 <title>Keeping Intermediate Files</title>
623 <indexterm><primary>intermediate files, saving</primary>
625 <indexterm><primary><literal>.hc</literal> files, saving</primary>
627 <indexterm><primary><literal>.s</literal> files, saving</primary>
631 <para>The following options are useful for keeping certain
632 intermediate files around, when normally GHC would throw these
633 away after compilation:</para>
637 <term><literal>-keep-hc-files</literal></term>
639 <primary><literal>-keep-hc-files</literal></primary>
642 <para>Keep intermediate <literal>.hc</literal> files when
643 doing <literal>.hs</literal>-to-<literal>.o</literal>
644 compilations via C (NOTE: <literal>.hc</literal> files
645 aren't generated when using the native code generator, you
646 may need to use <literal>-fvia-C</literal> to force them
647 to be produced).</para>
652 <term><literal>-keep-s-files</literal></term>
654 <primary><literal>-keep-s-files</literal></primary>
657 <para>Keep intermediate <literal>.s</literal> files.</para>
662 <term><literal>-keep-raw-s-files</literal></term>
664 <primary><literal>-keep-raw-s-files</literal></primary>
667 <para>Keep intermediate <literal>.raw-s</literal> files.
668 These are the direct output from the C compiler, before
669 GHC does “assembly mangling” to produce the
670 <literal>.s</literal> file. Again, these are not produced
671 when using the native code generator.</para>
676 <term><literal>-keep-tmp-files</literal></term>
678 <primary><literal>-keep-tmp-files</literal></primary>
681 <primary>temporary files</primary>
682 <secondary>keeping</secondary>
685 <para>Instructs the GHC driver not to delete any of its
686 temporary files, which it normally keeps in
687 <literal>/tmp</literal> (or possibly elsewhere; see <xref
688 linkend="temp-files">). Running GHC with
689 <literal>-v</literal> will show you what temporary files
690 were generated along the way.</para>
696 <sect2 id="temp-files">
697 <title>Redirecting temporary files</title>
700 <primary>temporary files</primary>
701 <secondary>redirecting</secondary>
706 <term><literal>-tmpdir</literal></term>
707 <indexterm><primary><literal>-tmpdir</literal></primary></indexterm>
709 <para>If you have trouble because of running out of space
710 in <filename>/tmp</filename> (or wherever your
711 installation thinks temporary files should go), you may
712 use the <option>-tmpdir
713 <dir></option><IndexTerm><Primary>-tmpdir
714 <dir> option</Primary></IndexTerm> option to specify
715 an alternate directory. For example, <option>-tmpdir
716 .</option> says to put temporary files in the current
717 working directory.</para>
719 <para>Alternatively, use your <Constant>TMPDIR</Constant>
720 environment variable.<IndexTerm><Primary>TMPDIR
721 environment variable</Primary></IndexTerm> Set it to the
722 name of the directory where temporary files should be put.
723 GCC and other programs will honour the
724 <Constant>TMPDIR</Constant> variable as well.</para>
726 <para>Even better idea: Set the
727 <Constant>DEFAULT_TMPDIR</Constant> make variable when
728 building GHC, and never worry about
729 <Constant>TMPDIR</Constant> again. (see the build
730 documentation).</para>
738 <sect1 id="options-sanity">
739 <title>Warnings and sanity-checking</title>
741 <indexterm><primary>sanity-checking options</primary></indexterm>
742 <indexterm><primary>warnings</primary></indexterm>
745 <para>GHC has a number of options that select which types of
746 non-fatal error messages, otherwise known as warnings, can be
747 generated during compilation. By default, you get a standard set
748 of warnings which are generally likely to indicate bugs in your
750 <option>-fwarn-overlpapping-patterns</option>,
751 <option>-fwarn-deprecations</option>,
752 <option>-fwarn-duplicate-exports</option>,
753 <option>-fwarn-missing-fields</option>, and
754 <option>-fwarn-missing-methods</option>. The following flags are
755 simple ways to select standard “packages” of warnings:
761 <term><option>-W</option>:</term>
763 <IndexTerm><Primary>-W option</Primary></IndexTerm>
764 <para>Provides the standard warnings plus
765 <option>-fwarn-incomplete-patterns</option>,
766 <option>-fwarn-unused-matches</option>,
767 <option>-fwarn-unused-imports</option> and
768 <option>-fwarn-unused-binds</option>.</para>
773 <term><option>-w</option>:</term>
775 <IndexTerm><Primary>-w option</Primary></IndexTerm>
776 <para>Turns off all warnings, including the standard ones.</para>
781 <term><option>-Wall</option>:</term>
783 <indexterm><primary>-Wall option</primary></indexterm>
784 <para>Turns on all warning options.</para>
790 <para>The full set of warning options is described below. To turn
791 off any warning, simply give the corresponding
792 <option>-fno-warn-...</option> option on the command line.</para>
797 <term><option>-fwarn-deprecations</option>:</term>
799 <indexterm><primary><option>-fwarn-deprecations</option></primary>
801 <indexterm><primary>deprecations</primary></indexterm>
802 <para>Causes a warning to be emitted when a deprecated
803 function or type is used. Entities can be marked as
804 deprecated using a pragma, see <xref
805 linkend="deprecated-pragma">.</para>
810 <term><option>-fwarn-duplicate-exports</option>:</term>
812 <indexterm><primary>-fwarn-duplicate-exports option</primary></indexterm>
813 <indexterm><primary>duplicate exports, warning</primary></indexterm>
814 <indexterm><primary>export lists, duplicates</primary></indexterm>
816 <para>Have the compiler warn about duplicate entries in
817 export lists. This is useful information if you maintain
818 large export lists, and want to avoid the continued export
819 of a definition after you've deleted (one) mention of it in
820 the export list.</para>
822 <para>This option is on by default.</para>
827 <term><option>-fwarn-hi-shadowing</option>:</term>
829 <indexterm><primary>-fwarn-hi-shadowing option</primary></indexterm>
830 <indexterm><primary>shadowing</primary>
831 <secondary>interface files</secondary></indexterm>
833 <para>Causes the compiler to emit a warning when a module or
834 interface file in the current directory is shadowing one
835 with the same module name in a library or other
841 <term><option>-fwarn-incomplete-patterns</option>:</term>
843 <indexterm><primary>-fwarn-incomplete-patterns option</primary></indexterm>
844 <indexterm><primary>incomplete patterns, warning</primary></indexterm>
845 <indexterm><primary>patterns, incomplete</primary></indexterm>
847 <para>Similarly for incomplete patterns, the function
848 <function>g</function> below will fail when applied to
849 non-empty lists, so the compiler will emit a warning about
850 this when <option>-fwarn-incomplete-patterns</option> is
857 <para>This option isn't enabled be default because it can be
858 a bit noisy, and it doesn't always indicate a bug in the
859 program. However, it's generally considered good practice
860 to cover all the cases in your functions.</para>
865 <term><option>-fwarn-missing-fields</option>:</term>
867 <indexterm><primary>-fwarn-missing-fields option</primary></indexterm>
868 <indexterm><primary>missing fields, warning</primary></indexterm>
869 <indexterm><primary>fields, missing</primary></indexterm>
871 <para>This option is on by default, and warns you whenever
872 the construction of a labelled field constructor isn't
873 complete, missing initializers for one or more fields. While
874 not an error (the missing fields are initialised with
875 bottoms), it is often an indication of a programmer error.</para>
880 <term><option>-fwarn-missing-methods</option>:</term>
882 <indexterm><primary>-fwarn-missing-methods option</primary></indexterm>
883 <indexterm><primary>missing methods, warning</primary></indexterm>
884 <indexterm><primary>methods, missing</primary></indexterm>
886 <para>This option is on by default, and warns you whenever
887 an instance declaration is missing one or more methods, and
888 the corresponding class declaration has no default
889 declaration for them.</para>
894 <term><option>-fwarn-missing-signatures</option>:</term>
896 <indexterm><primary>-fwarn-missing-signatures option</primary></indexterm>
897 <indexterm><primary>type signatures, missing</primary></indexterm>
899 <para>If you would like GHC to check that every top-level
900 function/value has a type signature, use the
901 <option>-fwarn-missing-signatures</option> option. This
902 option is off by default.</para>
907 <term><option>-fwarn-name-shadowing</option>:</term>
909 <indexterm><primary>-fwarn-name-shadowing option</primary></indexterm>
910 <indexterm><primary>shadowing, warning</primary></indexterm>
912 <para>This option causes a warning to be emitted whenever an
913 inner-scope value has the same name as an outer-scope value,
914 i.e. the inner value shadows the outer one. This can catch
915 typographical errors that turn into hard-to-find bugs, e.g.,
916 in the inadvertent cyclic definition <literal>let x = ... x
917 ... in</literal>.</para>
919 <para>Consequently, this option does
920 <emphasis>will</emphasis> complain about cyclic recursive
926 <term><option>-fwarn-overlapping-patterns</option>:</term>
927 <indexterm><primary>-fwarn-overlapping-patterns option</primary></indexterm>
928 <indexterm><primary>overlapping patterns, warning</primary></indexterm>
929 <indexterm><primary>patterns, overlapping</primary></indexterm>
931 <para>By default, the compiler will warn you if a set of
932 patterns are overlapping, i.e.,</para>
935 f :: String -> Int
941 <para>where the last pattern match in <Function>f</Function>
942 won't ever be reached, as the second pattern overlaps
943 it. More often than not, redundant patterns is a programmer
944 mistake/error, so this option is enabled by default.</para>
949 <term><option>-fwarn-simple-patterns</option>:</term>
951 <indexterm><primary><option>-fwarn-simple-patterns</option></primary>
953 <para>Causes the compiler to warn about lambda-bound
954 patterns that can fail, eg. <literal>\(x:xs)->...</literal>.
955 Normally, these aren't treated as incomplete patterns by
956 <option>-fwarn-incomplete-patterns</option>.</para>
961 <term><option>-fwarn-type-defaults</option>:</term>
963 <indexterm><primary>-fwarn-type-defaults option</primary></indexterm>
964 <indexterm><primary>defaulting mechanism, warning</primary></indexterm>
965 <para>Have the compiler warn/inform you where in your source
966 the Haskell defaulting mechanism for numeric types kicks
967 in. This is useful information when converting code from a
968 context that assumed one default into one with another,
969 e.g., the `default default' for Haskell 1.4 caused the
970 otherwise unconstrained value <Constant>1</Constant> to be
971 given the type <literal>Int</literal>, whereas Haskell 98
972 defaults it to <literal>Integer</literal>. This may lead to
973 differences in performance and behaviour, hence the
974 usefulness of being non-silent about this.</para>
976 <para>This warning is off by default.</para>
981 <term><option>-fwarn-unused-binds</option>:</term>
983 <indexterm><primary>-fwarn-unused-binds option</primary></indexterm>
984 <indexterm><primary>unused binds, warning</primary></indexterm>
985 <indexterm><primary>binds, unused</primary></indexterm>
986 <para>Report any function definitions (and local bindings)
987 which are unused. For top-level functions, the warning is
988 only given if the binding is not exported.</para>
993 <term><option>-fwarn-unused-imports</option>:</term>
995 <indexterm><primary>-fwarn-unused-imports option</primary></indexterm>
996 <indexterm><primary>unused imports, warning</primary></indexterm>
997 <indexterm><primary>imports, unused</primary></indexterm>
999 <para>Report any objects that are explicitly imported but
1005 <term><option>-fwarn-unused-matches</option>:</term>
1007 <indexterm><primary>-fwarn-unused-matches option</primary></indexterm>
1008 <indexterm><primary>unused matches, warning</primary></indexterm>
1009 <indexterm><primary>matches, unused</primary></indexterm>
1011 <para>Report all unused variables which arise from pattern
1012 matches, including patterns consisting of a single variable.
1013 For instance <literal>f x y = []</literal> would report
1014 <VarName>x</VarName> and <VarName>y</VarName> as unused. To
1015 eliminate the warning, all unused variables can be replaced
1016 with wildcards.</para>
1022 <para>If you're feeling really paranoid, the
1023 <option>-dcore-lint</option> option<indexterm><primary>-dcore-lint
1024 option</primary></indexterm> is a good choice. It turns on
1025 heavyweight intra-pass sanity-checking within GHC. (It checks
1026 GHC's sanity, not yours.)</para>
1033 <sect1 id="options-optimise">
1034 <title>Optimisation (code improvement)</title>
1036 <indexterm><primary>optimisation</primary></indexterm>
1037 <indexterm><primary>improvement, code</primary></indexterm>
1039 <para>The <option>-O*</option> options specify convenient
1040 “packages” of optimisation flags; the
1041 <option>-f*</option> options described later on specify
1042 <emphasis>individual</emphasis> optimisations to be turned on/off;
1043 the <option>-m*</option> options specify
1044 <emphasis>machine-specific</emphasis> optimisations to be turned
1047 <sect2 id="optimise-pkgs">
1048 <title><option>-O*</option>: convenient “packages” of optimisation flags.</title>
1050 <para>There are <emphasis>many</emphasis> options that affect
1051 the quality of code produced by GHC. Most people only have a
1052 general goal, something like “Compile quickly” or
1053 “Make my program run like greased lightning.” The
1054 following “packages” of optimisations (or lack
1055 thereof) should suffice.</para>
1057 <para>Once you choose a <option>-O*</option>
1058 “package,” stick with it—don't chop and
1059 change. Modules' interfaces <emphasis>will</emphasis> change
1060 with a shift to a new <option>-O*</option> option, and you may
1061 have to recompile a large chunk of all importing modules before
1062 your program can again be run safely (see <XRef
1063 LinkEnd="recomp">).</para>
1068 <term>No <option>-O*</option>-type option specified:</term>
1069 <indexterm><primary>-O* not specified</primary></indexterm>
1071 <para>This is taken to mean: “Please compile
1072 quickly; I'm not over-bothered about compiled-code
1073 quality.” So, for example: <command>ghc -c
1074 Foo.hs</command></para>
1079 <term><option>-O0</option>:</term>
1080 <indexterm><primary><option>-O0</option></primary></indexterm>
1082 <para>Means “turn off all optimisation”,
1083 reverting to the same settings as if no
1084 <option>-O</option> options had been specified. Saying
1085 <option>-O0</option> can be useful if
1086 eg. <command>make</command> has inserted a
1087 <option>-O</option> on the command line already.</para>
1092 <term><option>-O</option> or <option>-O1</option>:</term>
1093 <indexterm><primary>-O option</primary></indexterm>
1094 <indexterm><primary>-O1 option</primary></indexterm>
1095 <indexterm><primary>optimise</primary><secondary>normally</secondary></indexterm>
1097 <para>Means: “Generate good-quality code without
1098 taking too long about it.” Thus, for example:
1099 <command>ghc -c -O Main.lhs</command></para>
1104 <term><option>-O2</option>:</term>
1105 <indexterm><primary>-O2 option</primary></indexterm>
1106 <indexterm><primary>optimise</primary><secondary>aggressively</secondary></indexterm>
1108 <para>Means: “Apply every non-dangerous
1109 optimisation, even if it means significantly longer
1110 compile times.”</para>
1112 <para>The avoided “dangerous” optimisations
1113 are those that can make runtime or space
1114 <emphasis>worse</emphasis> if you're unlucky. They are
1115 normally turned on or off individually.</para>
1117 <para>At the moment, <option>-O2</option> is
1118 <emphasis>unlikely</emphasis> to produce better code than
1119 <option>-O</option>.</para>
1124 <term><option>-O2-for-C</option>:</term>
1125 <indexterm><primary>-O2-for-C option</primary></indexterm>
1126 <indexterm><primary>gcc, invoking with -O2</primary></indexterm>
1128 <para>Says to run GCC with <option>-O2</option>, which may
1129 be worth a few percent in execution speed. Don't forget
1130 <option>-fvia-C</option>, lest you use the native-code
1131 generator and bypass GCC altogether!</para>
1136 <term><option>-Ofile <file></option>:</term>
1137 <indexterm><primary>-Ofile <file> option</primary></indexterm>
1138 <indexterm><primary>optimising, customised</primary></indexterm>
1140 <para>(NOTE: not supported yet in GHC 5.x. Please ask if
1141 you're interested in this.)</para>
1143 <para>For those who need <emphasis>absolute</emphasis>
1144 control over <emphasis>exactly</emphasis> what options are
1145 used (e.g., compiler writers, sometimes :-), a list of
1146 options can be put in a file and then slurped in with
1147 <option>-Ofile</option>.</para>
1149 <para>In that file, comments are of the
1150 <literal>#</literal>-to-end-of-line variety; blank
1151 lines and most whitespace is ignored.</para>
1153 <para>Please ask if you are baffled and would like an
1154 example of <option>-Ofile</option>!</para>
1159 <para>We don't use a <option>-O*</option> flag for day-to-day
1160 work. We use <option>-O</option> to get respectable speed;
1161 e.g., when we want to measure something. When we want to go for
1162 broke, we tend to use <option>-O -fvia-C -O2-for-C</option> (and
1163 we go for lots of coffee breaks).</para>
1165 <para>The easiest way to see what <option>-O</option> (etc.)
1166 “really mean” is to run with <option>-v</option>,
1167 then stand back in amazement.</para>
1170 <sect2 id="options-f">
1171 <title><option>-f*</option>: platform-independent flags</title>
1173 <indexterm><primary>-f* options (GHC)</primary></indexterm>
1174 <indexterm><primary>-fno-* options (GHC)</primary></indexterm>
1176 <para>These flags turn on and off individual optimisations.
1177 They are normally set via the <option>-O</option> options
1178 described above, and as such, you shouldn't need to set any of
1179 them explicitly (indeed, doing so could lead to unexpected
1180 results). However, there are one or two that may be of
1185 <term><option>-fexcess-precision</option>:</term>
1187 <indexterm><primary><option>-fexcess-precision</option></primary></indexterm>
1188 <para>When this option is given, intermediate floating
1189 point values can have a <emphasis>greater</emphasis>
1190 precision/range than the final type. Generally this is a
1191 good thing, but some programs may rely on the exact
1193 <literal>Float</literal>/<literal>Double</literal> values
1194 and should not use this option for their compilation.</para>
1199 <term><option>-fignore-asserts</option>:</term>
1201 <indexterm><primary><option>-fignore-asserts</option></primary></indexterm>
1202 <para>Causes GHC to ignore uses of the function
1203 <literal>Exception.assert</literal> in source code (in
1204 other words, rewriting <literal>Exception.assert p
1205 e</literal> to <literal>e</literal> (see <xref
1206 linkend="sec-assertions">). This flag is turned on by
1207 <option>-O</option>.
1213 <term><option>-fno-strictness</option></term>
1214 <indexterm><primary><option>-fno-strictness</option></primary>
1217 <para>Turns off the strictness analyser; sometimes it eats
1218 too many cycles.</para>
1223 <term><option>-fno-cpr-analyse</option></term>
1224 <indexterm><primary><option>-fno-cpr-analyse</option></primary>
1227 <para>Turns off the CPR (constructed product result)
1228 analysis; it is somewhat experimental.</para>
1233 <term><option>-funbox-strict-fields</option>:</term>
1235 <indexterm><primary><option>-funbox-strict-fields</option></primary></indexterm>
1236 <indexterm><primary>strict constructor fields</primary></indexterm>
1237 <indexterm><primary>constructor fields, strict</primary></indexterm>
1239 <para>This option causes all constructor fields which are
1240 marked strict (i.e. “!”) to be unboxed or
1241 unpacked if possible. For example:</para>
1244 data T = T !Float !Float
1247 <para>will create a constructor <literal>T</literal>
1248 containing two unboxed floats if the
1249 <option>-funbox-strict-fields</option> flag is given.
1250 This may not always be an optimisation: if the
1251 <Function>T</Function> constructor is scrutinised and the
1252 floats passed to a non-strict function for example, they
1253 will have to be reboxed (this is done automatically by the
1256 <para>This option should only be used in conjunction with
1257 <option>-O</option>, in order to expose unfoldings to the
1258 compiler so the reboxing can be removed as often as
1259 possible. For example:</para>
1263 f (T f1 f2) = f1 + f2
1266 <para>The compiler will avoid reboxing
1267 <Function>f1</Function> and <Function>f2</Function> by
1268 inlining <Function>+</Function> on floats, but only when
1269 <option>-O</option> is on.</para>
1271 <para>Any single-constructor data is eligible for
1272 unpacking; for example</para>
1275 data T = T !(Int,Int)
1278 <para>will store the two <literal>Int</literal>s directly
1279 in the <Function>T</Function> constructor, by flattening
1280 the pair. Multi-level unpacking is also supported:</para>
1284 data S = S !Int !Int
1287 <para>will store two unboxed <literal>Int#</literal>s
1288 directly in the <Function>T</Function> constructor.</para>
1293 <term><option>-funfolding-update-in-place<n></option></term>
1294 <indexterm><primary><option>-funfolding-update-in-place</option></primary></indexterm>
1296 <para>Switches on an experimental "optimisation".
1297 Switching it on makes the compiler a little keener to
1298 inline a function that returns a constructor, if the
1299 context is that of a thunk.
1303 If we inlined plusInt we might get an opportunity to use
1304 update-in-place for the thunk 'x'.</para>
1309 <term><option>-funfolding-creation-threshold<n></option>:</term>
1311 <indexterm><primary><option>-funfolding-creation-threshold</option></primary></indexterm>
1312 <indexterm><primary>inlining, controlling</primary></indexterm>
1313 <indexterm><primary>unfolding, controlling</primary></indexterm>
1315 <para>(Default: 45) Governs the maximum size that GHC will
1316 allow a function unfolding to be. (An unfolding has a
1317 “size” that reflects the cost in terms of
1318 “code bloat” of expanding that unfolding at
1319 at a call site. A bigger function would be assigned a
1320 bigger cost.) </para>
1322 <para> Consequences: (a) nothing larger than this will be
1323 inlined (unless it has an INLINE pragma); (b) nothing
1324 larger than this will be spewed into an interface
1328 <para> Increasing this figure is more likely to result in longer
1329 compile times than faster code. The next option is more
1335 <term><option>-funfolding-use-threshold<n></option>:</term>
1337 <indexterm><primary><option>-funfolding-use-threshold</option></primary></indexterm>
1338 <indexterm><primary>inlining, controlling</primary></indexterm>
1339 <indexterm><primary>unfolding, controlling</primary></indexterm>
1341 <para>(Default: 8) This is the magic cut-off figure for
1342 unfolding: below this size, a function definition will be
1343 unfolded at the call-site, any bigger and it won't. The
1344 size computed for a function depends on two things: the
1345 actual size of the expression minus any discounts that
1346 apply (see <option>-funfolding-con-discount</option>).</para>
1357 <Sect1 id="sec-using-concurrent">
1358 <title>Using Concurrent Haskell</title>
1361 <indexterm><primary>Concurrent Haskell—use</primary></indexterm>
1365 GHC (as of version 4.00) supports Concurrent Haskell by default,
1366 without requiring a special option or libraries compiled in a certain
1367 way. To get access to the support libraries for Concurrent Haskell
1368 (i.e. <literal>Concurrent</literal> and friends), use the
1369 <option>-package concurrent</option> option.
1373 Three RTS options are provided for modifying the behaviour of the
1374 threaded runtime system. See the descriptions of
1375 <option>-C[<us>]</option>, <option>-q</option>, and
1376 <option>-t<num></option> in <XRef LinkEnd="parallel-rts-opts">.
1380 Concurrent Haskell is described in more detail in <XRef
1381 LinkEnd="sec-Concurrent">.
1386 <Sect1 id="sec-using-parallel">
1387 <title>Using Parallel Haskell</title>
1390 <indexterm><primary>Parallel Haskell—use</primary></indexterm>
1394 [You won't be able to execute parallel Haskell programs unless PVM3
1395 (Parallel Virtual Machine, version 3) is installed at your site.]
1399 To compile a Haskell program for parallel execution under PVM, use the
1400 <Option>-parallel</Option> option,<IndexTerm><Primary>-parallel
1401 option</Primary></IndexTerm> both when compiling <Emphasis>and
1402 linking</Emphasis>. You will probably want to <Literal>import
1403 Parallel</Literal> into your Haskell modules.
1407 To run your parallel program, once PVM is going, just invoke it
1408 “as normal”. The main extra RTS option is
1409 <Option>-qp<n></Option>, to say how many PVM
1410 “processors” your program to run on. (For more details of
1411 all relevant RTS options, please see <XRef
1412 LinkEnd="parallel-rts-opts">.)
1416 In truth, running Parallel Haskell programs and getting information
1417 out of them (e.g., parallelism profiles) is a battle with the vagaries of
1418 PVM, detailed in the following sections.
1421 <Sect2 id="pvm-dummies">
1422 <Title>Dummy's guide to using PVM</Title>
1425 <indexterm><primary>PVM, how to use</primary></indexterm>
1426 <indexterm><primary>Parallel Haskell—PVM use</primary></indexterm>
1427 Before you can run a parallel program under PVM, you must set the
1428 required environment variables (PVM's idea, not ours); something like,
1429 probably in your <filename>.cshrc</filename> or equivalent:
1432 setenv PVM_ROOT /wherever/you/put/it
1433 setenv PVM_ARCH `$PVM_ROOT/lib/pvmgetarch`
1434 setenv PVM_DPATH $PVM_ROOT/lib/pvmd
1440 Creating and/or controlling your “parallel machine” is a purely-PVM
1441 business; nothing specific to Parallel Haskell. The following paragraphs
1442 describe how to configure your parallel machine interactively.
1446 If you use parallel Haskell regularly on the same machine configuration it
1447 is a good idea to maintain a file with all machine names and to make the
1448 environment variable PVM_HOST_FILE point to this file. Then you can avoid
1449 the interactive operations described below by just saying
1457 You use the <Command>pvm</Command><IndexTerm><Primary>pvm command</Primary></IndexTerm> command to start PVM on your
1458 machine. You can then do various things to control/monitor your
1459 “parallel machine;” the most useful being:
1465 <ColSpec Align="Left">
1469 <entry><KeyCombo><KeyCap>Control</KeyCap><KeyCap>D</KeyCap></KeyCombo></entry>
1470 <entry>exit <command>pvm</command>, leaving it running</entry>
1474 <entry><command>halt</command></entry>
1475 <entry>kill off this “parallel machine” & exit</entry>
1479 <entry><command>add <host></command></entry>
1480 <entry>add <command><host></command> as a processor</entry>
1484 <entry><command>delete <host></command></entry>
1485 <entry>delete <command><host></command></entry>
1489 <entry><command>reset</command></entry>
1490 <entry>kill what's going, but leave PVM up</entry>
1494 <entry><command>conf</command></entry>
1495 <entry>list the current configuration</entry>
1499 <entry><command>ps</command></entry>
1500 <entry>report processes' status</entry>
1504 <entry><command>pstat <pid></command></entry>
1505 <entry>status of a particular process</entry>
1514 The PVM documentation can tell you much, much more about <command>pvm</command>!
1519 <Sect2 id="par-profiles">
1520 <Title>Parallelism profiles</Title>
1523 <indexterm><primary>parallelism profiles</primary></indexterm>
1524 <indexterm><primary>profiles, parallelism</primary></indexterm>
1525 <indexterm><primary>visualisation tools</primary></indexterm>
1529 With Parallel Haskell programs, we usually don't care about the
1530 results—only with “how parallel” it was! We want pretty pictures.
1534 Parallelism profiles (à la <Command>hbcpp</Command>) can be generated with the
1535 <Option>-qP</Option><IndexTerm><Primary>-qP RTS option (concurrent, parallel)</Primary></IndexTerm> RTS option. The
1536 per-processor profiling info is dumped into files named
1537 <Filename><full-path><program>.gr</Filename>. These are then munged into a PostScript picture,
1538 which you can then display. For example, to run your program
1539 <Filename>a.out</Filename> on 8 processors, then view the parallelism profile, do:
1545 <prompt>$</prompt> ./a.out +RTS -qP -qp8
1546 <prompt>$</prompt> grs2gr *.???.gr > temp.gr # combine the 8 .gr files into one
1547 <prompt>$</prompt> gr2ps -O temp.gr # cvt to .ps; output in temp.ps
1548 <prompt>$</prompt> ghostview -seascape temp.ps # look at it!
1554 The scripts for processing the parallelism profiles are distributed
1555 in <filename>ghc/utils/parallel/</filename>.
1561 <Title>Other useful info about running parallel programs</Title>
1564 The “garbage-collection statistics” RTS options can be useful for
1565 seeing what parallel programs are doing. If you do either
1566 <Option>+RTS -Sstderr</Option><IndexTerm><Primary>-Sstderr RTS option</Primary></IndexTerm> or <Option>+RTS -sstderr</Option>, then
1567 you'll get mutator, garbage-collection, etc., times on standard
1568 error. The standard error of all PE's other than the `main thread'
1569 appears in <filename>/tmp/pvml.nnn</filename>, courtesy of PVM.
1573 Whether doing <option>+RTS -Sstderr</option> or not, a handy way to watch
1574 what's happening overall is: <command>tail -f /tmp/pvml.nnn</command>.
1579 <Sect2 id="parallel-rts-opts">
1580 <title>RTS options for Concurrent/Parallel Haskell
1584 <indexterm><primary>RTS options, concurrent</primary></indexterm>
1585 <indexterm><primary>RTS options, parallel</primary></indexterm>
1586 <indexterm><primary>Concurrent Haskell—RTS options</primary></indexterm>
1587 <indexterm><primary>Parallel Haskell—RTS options</primary></indexterm>
1591 Besides the usual runtime system (RTS) options
1592 (<XRef LinkEnd="runtime-control">), there are a few options particularly
1593 for concurrent/parallel execution.
1600 <Term><Option>-qp<N></Option>:</Term>
1603 <IndexTerm><Primary>-qp<N> RTS option</Primary></IndexTerm>
1604 (PARALLEL ONLY) Use <Literal><N></Literal> PVM processors to run this program;
1610 <term><option>-C[<us>]</option>:</term>
1613 <indexterm><primary>-C<us> RTS option</primary></indexterm> Sets
1614 the context switch interval to <literal><s></literal> seconds.
1615 A context switch will occur at the next heap block allocation after
1616 the timer expires (a heap block allocation occurs every 4k of
1617 allocation). With <option>-C0</option> or <option>-C</option>,
1618 context switches will occur as often as possible (at every heap block
1619 allocation). By default, context switches occur every 20ms
1620 milliseconds. Note that GHC's internal timer ticks every 20ms, and
1621 the context switch timer is always a multiple of this timer, so 20ms
1622 is the maximum granularity available for timed context switches.
1627 <term><option>-q[v]</option>:</term>
1630 <indexterm><primary>-q RTS option</primary></indexterm>
1631 (PARALLEL ONLY) Produce a quasi-parallel profile of thread activity,
1632 in the file <FIlename><program>.qp</FIlename>. In the style of <command>hbcpp</command>, this profile
1633 records the movement of threads between the green (runnable) and red
1634 (blocked) queues. If you specify the verbose suboption (<option>-qv</option>), the
1635 green queue is split into green (for the currently running thread
1636 only) and amber (for other runnable threads). We do not recommend
1637 that you use the verbose suboption if you are planning to use the
1638 <Command>hbcpp</Command> profiling tools or if you are context switching at every heap
1639 check (with <Option>-C</Option>).
1645 <Term><Option>-qt<num></Option>:</Term>
1648 <IndexTerm><Primary>-qt<num> RTS option</Primary></IndexTerm>
1649 (PARALLEL ONLY) Limit the thread pool size, i.e. the number of concurrent
1650 threads per processor to <Literal><num></Literal>. The default is
1651 32. Each thread requires slightly over 1K <Emphasis>words</Emphasis> in
1652 the heap for thread state and stack objects. (For 32-bit machines, this
1653 translates to 4K bytes, and for 64-bit machines, 8K bytes.)
1659 <Term><Option>-d</Option>:</Term>
1662 <IndexTerm><Primary>-d RTS option (parallel)</Primary></IndexTerm>
1663 (PARALLEL ONLY) Turn on debugging. It pops up one xterm (or GDB, or
1664 something…) per PVM processor. We use the standard <Command>debugger</Command>
1665 script that comes with PVM3, but we sometimes meddle with the
1666 <Command>debugger2</Command> script. We include ours in the GHC distribution,
1667 in <Filename>ghc/utils/pvm/</Filename>.
1673 <Term><Option>-qe<num></Option>:</Term>
1676 <IndexTerm><Primary>-qe<num> RTS option
1677 (parallel)</Primary></IndexTerm> (PARALLEL ONLY) Limit the spark pool size
1678 i.e. the number of pending sparks per processor to
1679 <Literal><num></Literal>. The default is 100. A larger number may be
1680 appropriate if your program generates large amounts of parallelism
1686 <Term><Option>-qQ<num></Option>:</Term>
1689 <IndexTerm><Primary>-qQ<num> RTS option (parallel)</Primary></IndexTerm>
1690 (PARALLEL ONLY) Set the size of packets transmitted between processors
1691 to <Literal><num></Literal>. The default is 1024 words. A larger number may be
1692 appropriate if your machine has a high communication cost relative to
1698 <Term><Option>-qh<num></Option>:</Term>
1701 <IndexTerm><Primary>-qh<num> RTS option (parallel)</Primary></IndexTerm>
1702 (PARALLEL ONLY) Select a packing scheme. Set the number of non-root thunks to pack in one packet to
1703 <num>-1 (0 means infinity). By default GUM uses full-subgraph
1704 packing, i.e. the entire subgraph with the requested closure as root is
1705 transmitted (provided it fits into one packet). Choosing a smaller value
1706 reduces the amount of pre-fetching of work done in GUM. This can be
1707 advantageous for improving data locality but it can also worsen the balance
1708 of the load in the system.
1713 <Term><Option>-qg<num></Option>:</Term>
1716 <IndexTerm><Primary>-qg<num> RTS option
1717 (parallel)</Primary></IndexTerm> (PARALLEL ONLY) Select a globalisation
1718 scheme. This option affects the
1719 generation of global addresses when transferring data. Global addresses are
1720 globally unique identifiers required to maintain sharing in the distributed
1721 graph structure. Currently this is a binary option. With <num>=0 full globalisation is used
1722 (default). This means a global address is generated for every closure that
1723 is transmitted. With <num>=1 a thunk-only globalisation scheme is
1724 used, which generated global address only for thunks. The latter case may
1725 lose sharing of data but has a reduced overhead in packing graph structures
1726 and maintaining internal tables of global addresses.
1737 <sect1 id="options-platform">
1738 <title>Platform-specific Flags</title>
1740 <indexterm><primary>-m* options</primary></indexterm>
1741 <indexterm><primary>platform-specific options</primary></indexterm>
1742 <indexterm><primary>machine-specific options</primary></indexterm>
1744 <para>Some flags only make sense for particular target
1750 <term><option>-mv8</option>:</term>
1752 <para>(SPARC machines)<indexterm><primary>-mv8 option (SPARC
1753 only)</primary></indexterm> Means to pass the like-named
1754 option to GCC; it says to use the Version 8 SPARC
1755 instructions, notably integer multiply and divide. The
1756 similiar <option>-m*</option> GCC options for SPARC also
1757 work, actually.</para>
1762 <term><option>-monly-[32]-regs</option>:</term>
1764 <para>(iX86 machines)<indexterm><primary>-monly-N-regs
1765 option (iX86 only)</primary></indexterm> GHC tries to
1766 “steal” four registers from GCC, for performance
1767 reasons; it almost always works. However, when GCC is
1768 compiling some modules with four stolen registers, it will
1769 crash, probably saying:
1772 Foo.hc:533: fixed or forbidden register was spilled.
1773 This may be due to a compiler bug or to impossible asm
1774 statements or clauses.
1777 Just give some registers back with
1778 <option>-monly-N-regs</option>. Try `3' first, then `2'.
1779 If `2' doesn't work, please report the bug to us.</para>
1793 ;;; Local Variables: ***
1795 ;;; sgml-parent-document: ("users_guide.sgml" "book" "chapter") ***