1 <?xml version="1.0" encoding="iso-8859-1"?>
2 <chapter id="profiling">
3 <title>Profiling</title>
4 <indexterm><primary>profiling</primary>
6 <indexterm><primary>cost-centre profiling</primary></indexterm>
8 <para> Glasgow Haskell comes with a time and space profiling
9 system. Its purpose is to help you improve your understanding of
10 your program's execution behaviour, so you can improve it.</para>
12 <para> Any comments, suggestions and/or improvements you have are
13 welcome. Recommended “profiling tricks” would be
14 especially cool! </para>
16 <para>Profiling a program is a three-step process:</para>
20 <para> Re-compile your program for profiling with the
21 <literal>-prof</literal> option, and probably one of the
22 <literal>-auto</literal> or <literal>-auto-all</literal>
23 options. These options are described in more detail in <xref
24 linkend="prof-compiler-options"/> </para>
25 <indexterm><primary><literal>-prof</literal></primary>
27 <indexterm><primary><literal>-auto</literal></primary>
29 <indexterm><primary><literal>-auto-all</literal></primary>
34 <para> Run your program with one of the profiling options, eg.
35 <literal>+RTS -p -RTS</literal>. This generates a file of
36 profiling information. Note that multi-processor execution
37 (e.g. <literal>+RTS -N2</literal>) is not supported while
39 <indexterm><primary><option>-p</option></primary><secondary>RTS
40 option</secondary></indexterm>
44 <para> Examine the generated profiling information, using one of
45 GHC's profiling tools. The tool to use will depend on the kind
46 of profiling information generated.</para>
51 <sect1 id="cost-centres">
52 <title>Cost centres and cost-centre stacks</title>
54 <para>GHC's profiling system assigns <firstterm>costs</firstterm>
55 to <firstterm>cost centres</firstterm>. A cost is simply the time
56 or space required to evaluate an expression. Cost centres are
57 program annotations around expressions; all costs incurred by the
58 annotated expression are assigned to the enclosing cost centre.
59 Furthermore, GHC will remember the stack of enclosing cost centres
60 for any given expression at run-time and generate a call-graph of
61 cost attributions.</para>
63 <para>Let's take a look at an example:</para>
66 main = print (nfib 25)
67 nfib n = if n < 2 then 1 else nfib (n-1) + nfib (n-2)
70 <para>Compile and run this program as follows:</para>
73 $ ghc -prof -auto-all -o Main Main.hs
79 <para>When a GHC-compiled program is run with the
80 <option>-p</option> RTS option, it generates a file called
81 <filename><prog>.prof</filename>. In this case, the file
82 will contain something like this:</para>
85 Fri May 12 14:06 2000 Time and Allocation Profiling Report (Final)
89 total time = 0.14 secs (7 ticks @ 20 ms)
90 total alloc = 8,741,204 bytes (excludes profiling overheads)
92 COST CENTRE MODULE %time %alloc
98 COST CENTRE MODULE entries %time %alloc %time %alloc
100 MAIN MAIN 0 0.0 0.0 100.0 100.0
101 main Main 0 0.0 0.0 0.0 0.0
102 CAF PrelHandle 3 0.0 0.0 0.0 0.0
103 CAF PrelAddr 1 0.0 0.0 0.0 0.0
104 CAF Main 6 0.0 0.0 100.0 100.0
105 main Main 1 0.0 0.0 100.0 100.0
106 nfib Main 242785 100.0 100.0 100.0 100.0
110 <para>The first part of the file gives the program name and
111 options, and the total time and total memory allocation measured
112 during the run of the program (note that the total memory
113 allocation figure isn't the same as the amount of
114 <emphasis>live</emphasis> memory needed by the program at any one
115 time; the latter can be determined using heap profiling, which we
116 will describe shortly).</para>
118 <para>The second part of the file is a break-down by cost centre
119 of the most costly functions in the program. In this case, there
120 was only one significant function in the program, namely
121 <function>nfib</function>, and it was responsible for 100%
122 of both the time and allocation costs of the program.</para>
124 <para>The third and final section of the file gives a profile
125 break-down by cost-centre stack. This is roughly a call-graph
126 profile of the program. In the example above, it is clear that
127 the costly call to <function>nfib</function> came from
128 <function>main</function>.</para>
130 <para>The time and allocation incurred by a given part of the
131 program is displayed in two ways: “individual”, which
132 are the costs incurred by the code covered by this cost centre
133 stack alone, and “inherited”, which includes the costs
134 incurred by all the children of this node.</para>
136 <para>The usefulness of cost-centre stacks is better demonstrated
137 by modifying the example slightly:</para>
140 main = print (f 25 + g 25)
142 g n = nfib (n `div` 2)
143 nfib n = if n < 2 then 1 else nfib (n-1) + nfib (n-2)
146 <para>Compile and run this program as before, and take a look at
147 the new profiling results:</para>
150 COST CENTRE MODULE scc %time %alloc %time %alloc
152 MAIN MAIN 0 0.0 0.0 100.0 100.0
153 main Main 0 0.0 0.0 0.0 0.0
154 CAF PrelHandle 3 0.0 0.0 0.0 0.0
155 CAF PrelAddr 1 0.0 0.0 0.0 0.0
156 CAF Main 9 0.0 0.0 100.0 100.0
157 main Main 1 0.0 0.0 100.0 100.0
158 g Main 1 0.0 0.0 0.0 0.2
159 nfib Main 465 0.0 0.2 0.0 0.2
160 f Main 1 0.0 0.0 100.0 99.8
161 nfib Main 242785 100.0 99.8 100.0 99.8
164 <para>Now although we had two calls to <function>nfib</function>
165 in the program, it is immediately clear that it was the call from
166 <function>f</function> which took all the time.</para>
168 <para>The actual meaning of the various columns in the output is:</para>
174 <para>The number of times this particular point in the call
175 graph was entered.</para>
180 <term>individual %time</term>
182 <para>The percentage of the total run time of the program
183 spent at this point in the call graph.</para>
188 <term>individual %alloc</term>
190 <para>The percentage of the total memory allocations
191 (excluding profiling overheads) of the program made by this
197 <term>inherited %time</term>
199 <para>The percentage of the total run time of the program
200 spent below this point in the call graph.</para>
205 <term>inherited %alloc</term>
207 <para>The percentage of the total memory allocations
208 (excluding profiling overheads) of the program made by this
209 call and all of its sub-calls.</para>
214 <para>In addition you can use the <option>-P</option> RTS option
215 <indexterm><primary><option>-P</option></primary></indexterm> to
216 get the following additional information:</para>
220 <term><literal>ticks</literal></term>
222 <para>The raw number of time “ticks” which were
223 attributed to this cost-centre; from this, we get the
224 <literal>%time</literal> figure mentioned
230 <term><literal>bytes</literal></term>
232 <para>Number of bytes allocated in the heap while in this
233 cost-centre; again, this is the raw number from which we get
234 the <literal>%alloc</literal> figure mentioned
240 <para>What about recursive functions, and mutually recursive
241 groups of functions? Where are the costs attributed? Well,
242 although GHC does keep information about which groups of functions
243 called each other recursively, this information isn't displayed in
244 the basic time and allocation profile, instead the call-graph is
245 flattened into a tree.</para>
247 <sect2><title>Inserting cost centres by hand</title>
249 <para>Cost centres are just program annotations. When you say
250 <option>-auto-all</option> to the compiler, it automatically
251 inserts a cost centre annotation around every top-level function
252 in your program, but you are entirely free to add the cost
253 centre annotations yourself.</para>
255 <para>The syntax of a cost centre annotation is</para>
258 {-# SCC "name" #-} <expression>
261 <para>where <literal>"name"</literal> is an arbitrary string,
262 that will become the name of your cost centre as it appears
263 in the profiling output, and
264 <literal><expression></literal> is any Haskell
265 expression. An <literal>SCC</literal> annotation extends as
266 far to the right as possible when parsing. (SCC stands for "Set
267 Cost Centre").</para>
269 <para>Here is an example of a program with a couple of SCCs:</para>
273 main = do let xs = {-# SCC "X" #-} [1..1000000]
274 let ys = {-# SCC "Y" #-} [1..2000000]
276 print $ last $ init xs
278 print $ last $ init ys
281 <para>which gives this heap profile when run:</para>
283 <!-- contentwidth/contentheight don't appear to have any effect
284 other than making the PS file generation work, rather than
285 falling over. The result seems to be broken PS on the page
287 <imagedata fileref="prof_scc" contentwidth="645px"
288 contentdepth="428px"/>
292 <sect2 id="prof-rules">
293 <title>Rules for attributing costs</title>
295 <para>The cost of evaluating any expression in your program is
296 attributed to a cost-centre stack using the following rules:</para>
300 <para>If the expression is part of the
301 <firstterm>one-off</firstterm> costs of evaluating the
302 enclosing top-level definition, then costs are attributed to
303 the stack of lexically enclosing <literal>SCC</literal>
304 annotations on top of the special <literal>CAF</literal>
309 <para>Otherwise, costs are attributed to the stack of
310 lexically-enclosing <literal>SCC</literal> annotations,
311 appended to the cost-centre stack in effect at the
312 <firstterm>call site</firstterm> of the current top-level
313 definition<footnote> <para>The call-site is just the place
314 in the source code which mentions the particular function or
315 variable.</para></footnote>. Notice that this is a recursive
320 <para>Time spent in foreign code (see <xref linkend="ffi"/>)
321 is always attributed to the cost centre in force at the
322 Haskell call-site of the foreign function.</para>
326 <para>What do we mean by one-off costs? Well, Haskell is a lazy
327 language, and certain expressions are only ever evaluated once.
328 For example, if we write:</para>
334 <para>then <varname>x</varname> will only be evaluated once (if
335 at all), and subsequent demands for <varname>x</varname> will
336 immediately get to see the cached result. The definition
337 <varname>x</varname> is called a CAF (Constant Applicative
338 Form), because it has no arguments.</para>
340 <para>For the purposes of profiling, we say that the expression
341 <literal>nfib 25</literal> belongs to the one-off costs of
342 evaluating <varname>x</varname>.</para>
344 <para>Since one-off costs aren't strictly speaking part of the
345 call-graph of the program, they are attributed to a special
346 top-level cost centre, <literal>CAF</literal>. There may be one
347 <literal>CAF</literal> cost centre for each module (the
348 default), or one for each top-level definition with any one-off
349 costs (this behaviour can be selected by giving GHC the
350 <option>-caf-all</option> flag).</para>
352 <indexterm><primary><literal>-caf-all</literal></primary>
355 <para>If you think you have a weird profile, or the call-graph
356 doesn't look like you expect it to, feel free to send it (and
357 your program) to us at
358 <email>glasgow-haskell-bugs@haskell.org</email>.</para>
362 <sect1 id="prof-compiler-options">
363 <title>Compiler options for profiling</title>
365 <indexterm><primary>profiling</primary><secondary>options</secondary></indexterm>
366 <indexterm><primary>options</primary><secondary>for profiling</secondary></indexterm>
371 <option>-prof</option>:
372 <indexterm><primary><option>-prof</option></primary></indexterm>
375 <para> To make use of the profiling system
376 <emphasis>all</emphasis> modules must be compiled and linked
377 with the <option>-prof</option> option. Any
378 <literal>SCC</literal> annotations you've put in your source
379 will spring to life.</para>
381 <para> Without a <option>-prof</option> option, your
382 <literal>SCC</literal>s are ignored; so you can compile
383 <literal>SCC</literal>-laden code without changing
389 <para>There are a few other profiling-related compilation options.
390 Use them <emphasis>in addition to</emphasis>
391 <option>-prof</option>. These do not have to be used consistently
392 for all modules in a program.</para>
397 <option>-auto</option>:
398 <indexterm><primary><option>-auto</option></primary></indexterm>
399 <indexterm><primary>cost centres</primary><secondary>automatically inserting</secondary></indexterm>
402 <para> GHC will automatically add
403 <function>_scc_</function> constructs for all
404 top-level, exported functions.</para>
410 <option>-auto-all</option>:
411 <indexterm><primary><option>-auto-all</option></primary></indexterm>
414 <para> <emphasis>All</emphasis> top-level functions,
415 exported or not, will be automatically
416 <function>_scc_</function>'d.</para>
422 <option>-caf-all</option>:
423 <indexterm><primary><option>-caf-all</option></primary></indexterm>
426 <para> The costs of all CAFs in a module are usually
427 attributed to one “big” CAF cost-centre. With
428 this option, all CAFs get their own cost-centre. An
429 “if all else fails” option…</para>
435 <option>-ignore-scc</option>:
436 <indexterm><primary><option>-ignore-scc</option></primary></indexterm>
439 <para>Ignore any <function>_scc_</function>
440 constructs, so a module which already has
441 <function>_scc_</function>s can be compiled
442 for profiling with the annotations ignored.</para>
450 <sect1 id="prof-time-options">
451 <title>Time and allocation profiling</title>
453 <para>To generate a time and allocation profile, give one of the
454 following RTS options to the compiled program when you run it (RTS
455 options should be enclosed between <literal>+RTS...-RTS</literal>
461 <option>-p</option> or <option>-P</option> or <option>-pa</option>:
462 <indexterm><primary><option>-p</option></primary></indexterm>
463 <indexterm><primary><option>-P</option></primary></indexterm>
464 <indexterm><primary><option>-pa</option></primary></indexterm>
465 <indexterm><primary>time profile</primary></indexterm>
468 <para>The <option>-p</option> option produces a standard
469 <emphasis>time profile</emphasis> report. It is written
471 <filename><replaceable>program</replaceable>.prof</filename>.</para>
473 <para>The <option>-P</option> option produces a more
474 detailed report containing the actual time and allocation
475 data as well. (Not used much.)</para>
477 <para>The <option>-pa</option> option produces the most detailed
478 report containing all cost centres in addition to the actual time
479 and allocation data.</para>
484 <term><option>-V<replaceable>secs</replaceable></option>
485 <indexterm><primary><option>-V</option></primary><secondary>RTS
486 option</secondary></indexterm></term>
488 <para>Sets the interval that the RTS clock ticks at, which is
489 also the sampling interval of the time and allocation profile.
490 The default is 0.02 second.</para>
497 <indexterm><primary><option>-xc</option></primary><secondary>RTS option</secondary></indexterm>
500 <para>This option makes use of the extra information
501 maintained by the cost-centre-stack profiler to provide
502 useful information about the location of runtime errors.
503 See <xref linkend="rts-options-debugging"/>.</para>
511 <sect1 id="prof-heap">
512 <title>Profiling memory usage</title>
514 <para>In addition to profiling the time and allocation behaviour
515 of your program, you can also generate a graph of its memory usage
516 over time. This is useful for detecting the causes of
517 <firstterm>space leaks</firstterm>, when your program holds on to
518 more memory at run-time that it needs to. Space leaks lead to
519 longer run-times due to heavy garbage collector activity, and may
520 even cause the program to run out of memory altogether.</para>
522 <para>To generate a heap profile from your program:</para>
526 <para>Compile the program for profiling (<xref
527 linkend="prof-compiler-options"/>).</para>
530 <para>Run it with one of the heap profiling options described
531 below (eg. <option>-hc</option> for a basic producer profile).
532 This generates the file
533 <filename><replaceable>prog</replaceable>.hp</filename>.</para>
536 <para>Run <command>hp2ps</command> to produce a Postscript
538 <filename><replaceable>prog</replaceable>.ps</filename>. The
539 <command>hp2ps</command> utility is described in detail in
540 <xref linkend="hp2ps"/>.</para>
543 <para>Display the heap profile using a postscript viewer such
544 as <application>Ghostview</application>, or print it out on a
545 Postscript-capable printer.</para>
549 <para>You might also want to take a look
550 at <ulink url="http://www.haskell.org/haskellwiki/Hp2any">hp2any</ulink>,
551 a more advanced suite of tools (not distributed with GHC) for
552 displaying heap profiles.</para>
554 <sect2 id="rts-options-heap-prof">
555 <title>RTS options for heap profiling</title>
557 <para>There are several different kinds of heap profile that can
558 be generated. All the different profile types yield a graph of
559 live heap against time, but they differ in how the live heap is
560 broken down into bands. The following RTS options select which
561 break-down to use:</para>
567 <indexterm><primary><option>-hc</option></primary><secondary>RTS option</secondary></indexterm>
570 <para>Breaks down the graph by the cost-centre stack which
571 produced the data.</para>
578 <indexterm><primary><option>-hm</option></primary><secondary>RTS option</secondary></indexterm>
581 <para>Break down the live heap by the module containing
582 the code which produced the data.</para>
589 <indexterm><primary><option>-hd</option></primary><secondary>RTS option</secondary></indexterm>
592 <para>Breaks down the graph by <firstterm>closure
593 description</firstterm>. For actual data, the description
594 is just the constructor name, for other closures it is a
595 compiler-generated string identifying the closure.</para>
602 <indexterm><primary><option>-hy</option></primary><secondary>RTS option</secondary></indexterm>
605 <para>Breaks down the graph by
606 <firstterm>type</firstterm>. For closures which have
607 function type or unknown/polymorphic type, the string will
608 represent an approximation to the actual type.</para>
615 <indexterm><primary><option>-hr</option></primary><secondary>RTS option</secondary></indexterm>
618 <para>Break down the graph by <firstterm>retainer
619 set</firstterm>. Retainer profiling is described in more
620 detail below (<xref linkend="retainer-prof"/>).</para>
627 <indexterm><primary><option>-hb</option></primary><secondary>RTS option</secondary></indexterm>
630 <para>Break down the graph by
631 <firstterm>biography</firstterm>. Biographical profiling
632 is described in more detail below (<xref
633 linkend="biography-prof"/>).</para>
638 <para>In addition, the profile can be restricted to heap data
639 which satisfies certain criteria - for example, you might want
640 to display a profile by type but only for data produced by a
641 certain module, or a profile by retainer for a certain type of
642 data. Restrictions are specified as follows:</para>
647 <option>-hc</option><replaceable>name</replaceable>,...
648 <indexterm><primary><option>-hc</option></primary><secondary>RTS option</secondary></indexterm>
651 <para>Restrict the profile to closures produced by
652 cost-centre stacks with one of the specified cost centres
659 <option>-hC</option><replaceable>name</replaceable>,...
660 <indexterm><primary><option>-hC</option></primary><secondary>RTS option</secondary></indexterm>
663 <para>Restrict the profile to closures produced by
664 cost-centre stacks with one of the specified cost centres
665 anywhere in the stack.</para>
671 <option>-hm</option><replaceable>module</replaceable>,...
672 <indexterm><primary><option>-hm</option></primary><secondary>RTS option</secondary></indexterm>
675 <para>Restrict the profile to closures produced by the
676 specified modules.</para>
682 <option>-hd</option><replaceable>desc</replaceable>,...
683 <indexterm><primary><option>-hd</option></primary><secondary>RTS option</secondary></indexterm>
686 <para>Restrict the profile to closures with the specified
687 description strings.</para>
693 <option>-hy</option><replaceable>type</replaceable>,...
694 <indexterm><primary><option>-hy</option></primary><secondary>RTS option</secondary></indexterm>
697 <para>Restrict the profile to closures with the specified
704 <option>-hr</option><replaceable>cc</replaceable>,...
705 <indexterm><primary><option>-hr</option></primary><secondary>RTS option</secondary></indexterm>
708 <para>Restrict the profile to closures with retainer sets
709 containing cost-centre stacks with one of the specified
710 cost centres at the top.</para>
716 <option>-hb</option><replaceable>bio</replaceable>,...
717 <indexterm><primary><option>-hb</option></primary><secondary>RTS option</secondary></indexterm>
720 <para>Restrict the profile to closures with one of the
721 specified biographies, where
722 <replaceable>bio</replaceable> is one of
723 <literal>lag</literal>, <literal>drag</literal>,
724 <literal>void</literal>, or <literal>use</literal>.</para>
729 <para>For example, the following options will generate a
730 retainer profile restricted to <literal>Branch</literal> and
731 <literal>Leaf</literal> constructors:</para>
734 <replaceable>prog</replaceable> +RTS -hr -hdBranch,Leaf
737 <para>There can only be one "break-down" option
738 (eg. <option>-hr</option> in the example above), but there is no
739 limit on the number of further restrictions that may be applied.
740 All the options may be combined, with one exception: GHC doesn't
741 currently support mixing the <option>-hr</option> and
742 <option>-hb</option> options.</para>
744 <para>There are three more options which relate to heap
750 <option>-i<replaceable>secs</replaceable></option>:
751 <indexterm><primary><option>-i</option></primary></indexterm>
754 <para>Set the profiling (sampling) interval to
755 <replaceable>secs</replaceable> seconds (the default is
756 0.1 second). Fractions are allowed: for example
757 <option>-i0.2</option> will get 5 samples per second.
758 This only affects heap profiling; time profiles are always
759 sampled with the frequency of the RTS clock. See
760 <xref linkend="prof-time-options"/> for changing that.</para>
767 <indexterm><primary><option>-xt</option></primary><secondary>RTS option</secondary></indexterm>
770 <para>Include the memory occupied by threads in a heap
771 profile. Each thread takes up a small area for its thread
772 state in addition to the space allocated for its stack
773 (stacks normally start small and then grow as
776 <para>This includes the main thread, so using
777 <option>-xt</option> is a good way to see how much stack
778 space the program is using.</para>
780 <para>Memory occupied by threads and their stacks is
781 labelled as “TSO” when displaying the profile
782 by closure description or type description.</para>
788 <option>-L<replaceable>num</replaceable></option>
789 <indexterm><primary><option>-L</option></primary><secondary>RTS option</secondary></indexterm>
793 Sets the maximum length of a cost-centre stack name in a
794 heap profile. Defaults to 25.
802 <sect2 id="retainer-prof">
803 <title>Retainer Profiling</title>
805 <para>Retainer profiling is designed to help answer questions
806 like <quote>why is this data being retained?</quote>. We start
807 by defining what we mean by a retainer:</para>
810 <para>A retainer is either the system stack, or an unevaluated
811 closure (thunk).</para>
814 <para>In particular, constructors are <emphasis>not</emphasis>
817 <para>An object B retains object A if (i) B is a retainer object and
818 (ii) object A can be reached by recursively following pointers
819 starting from object B, but not meeting any other retainer
820 objects on the way. Each live object is retained by one or more
821 retainer objects, collectively called its retainer set, or its
822 <firstterm>retainer set</firstterm>, or its
823 <firstterm>retainers</firstterm>.</para>
825 <para>When retainer profiling is requested by giving the program
826 the <option>-hr</option> option, a graph is generated which is
827 broken down by retainer set. A retainer set is displayed as a
828 set of cost-centre stacks; because this is usually too large to
829 fit on the profile graph, each retainer set is numbered and
830 shown abbreviated on the graph along with its number, and the
831 full list of retainer sets is dumped into the file
832 <filename><replaceable>prog</replaceable>.prof</filename>.</para>
834 <para>Retainer profiling requires multiple passes over the live
835 heap in order to discover the full retainer set for each
836 object, which can be quite slow. So we set a limit on the
837 maximum size of a retainer set, where all retainer sets larger
838 than the maximum retainer set size are replaced by the special
839 set <literal>MANY</literal>. The maximum set size defaults to 8
840 and can be altered with the <option>-R</option> RTS
845 <term><option>-R</option><replaceable>size</replaceable></term>
847 <para>Restrict the number of elements in a retainer set to
848 <replaceable>size</replaceable> (default 8).</para>
854 <title>Hints for using retainer profiling</title>
856 <para>The definition of retainers is designed to reflect a
857 common cause of space leaks: a large structure is retained by
858 an unevaluated computation, and will be released once the
859 computation is forced. A good example is looking up a value in
860 a finite map, where unless the lookup is forced in a timely
861 manner the unevaluated lookup will cause the whole mapping to
862 be retained. These kind of space leaks can often be
863 eliminated by forcing the relevant computations to be
864 performed eagerly, using <literal>seq</literal> or strictness
865 annotations on data constructor fields.</para>
867 <para>Often a particular data structure is being retained by a
868 chain of unevaluated closures, only the nearest of which will
869 be reported by retainer profiling - for example A retains B, B
870 retains C, and C retains a large structure. There might be a
871 large number of Bs but only a single A, so A is really the one
872 we're interested in eliminating. However, retainer profiling
873 will in this case report B as the retainer of the large
874 structure. To move further up the chain of retainers, we can
875 ask for another retainer profile but this time restrict the
876 profile to B objects, so we get a profile of the retainers of
880 <replaceable>prog</replaceable> +RTS -hr -hcB
883 <para>This trick isn't foolproof, because there might be other
884 B closures in the heap which aren't the retainers we are
885 interested in, but we've found this to be a useful technique
886 in most cases.</para>
890 <sect2 id="biography-prof">
891 <title>Biographical Profiling</title>
893 <para>A typical heap object may be in one of the following four
894 states at each point in its lifetime:</para>
898 <para>The <firstterm>lag</firstterm> stage, which is the
899 time between creation and the first use of the
903 <para>the <firstterm>use</firstterm> stage, which lasts from
904 the first use until the last use of the object, and</para>
907 <para>The <firstterm>drag</firstterm> stage, which lasts
908 from the final use until the last reference to the object
912 <para>An object which is never used is said to be in the
913 <firstterm>void</firstterm> state for its whole
918 <para>A biographical heap profile displays the portion of the
919 live heap in each of the four states listed above. Usually the
920 most interesting states are the void and drag states: live heap
921 in these states is more likely to be wasted space than heap in
922 the lag or use states.</para>
924 <para>It is also possible to break down the heap in one or more
925 of these states by a different criteria, by restricting a
926 profile by biography. For example, to show the portion of the
927 heap in the drag or void state by producer: </para>
930 <replaceable>prog</replaceable> +RTS -hc -hbdrag,void
933 <para>Once you know the producer or the type of the heap in the
934 drag or void states, the next step is usually to find the
938 <replaceable>prog</replaceable> +RTS -hr -hc<replaceable>cc</replaceable>...
941 <para>NOTE: this two stage process is required because GHC
942 cannot currently profile using both biographical and retainer
943 information simultaneously.</para>
946 <sect2 id="mem-residency">
947 <title>Actual memory residency</title>
949 <para>How does the heap residency reported by the heap profiler relate to
950 the actual memory residency of your program when you run it? You might
951 see a large discrepancy between the residency reported by the heap
952 profiler, and the residency reported by tools on your system
953 (eg. <literal>ps</literal> or <literal>top</literal> on Unix, or the
954 Task Manager on Windows). There are several reasons for this:</para>
958 <para>There is an overhead of profiling itself, which is subtracted
959 from the residency figures by the profiler. This overhead goes
960 away when compiling without profiling support, of course. The
961 space overhead is currently 2 extra
962 words per heap object, which probably results in
963 about a 30% overhead.</para>
967 <para>Garbage collection requires more memory than the actual
968 residency. The factor depends on the kind of garbage collection
969 algorithm in use: a major GC in the standard
970 generation copying collector will usually require 3L bytes of
971 memory, where L is the amount of live data. This is because by
972 default (see the <option>+RTS -F</option> option) we allow the old
973 generation to grow to twice its size (2L) before collecting it, and
974 we require additionally L bytes to copy the live data into. When
975 using compacting collection (see the <option>+RTS -c</option>
976 option), this is reduced to 2L, and can further be reduced by
977 tweaking the <option>-F</option> option. Also add the size of the
978 allocation area (currently a fixed 512Kb).</para>
982 <para>The stack isn't counted in the heap profile by default. See the
983 <option>+RTS -xt</option> option.</para>
987 <para>The program text itself, the C stack, any non-heap data (eg. data
988 allocated by foreign libraries, and data allocated by the RTS), and
989 <literal>mmap()</literal>'d memory are not counted in the heap profile.</para>
997 <title><command>hp2ps</command>––heap profile to PostScript</title>
999 <indexterm><primary><command>hp2ps</command></primary></indexterm>
1000 <indexterm><primary>heap profiles</primary></indexterm>
1001 <indexterm><primary>postscript, from heap profiles</primary></indexterm>
1002 <indexterm><primary><option>-h<break-down></option></primary></indexterm>
1007 hp2ps [flags] [<file>[.hp]]
1011 <command>hp2ps</command><indexterm><primary>hp2ps
1012 program</primary></indexterm> converts a heap profile as produced
1013 by the <option>-h<break-down></option> runtime option into a
1014 PostScript graph of the heap profile. By convention, the file to
1015 be processed by <command>hp2ps</command> has a
1016 <filename>.hp</filename> extension. The PostScript output is
1017 written to <filename><file>@.ps</filename>. If
1018 <filename><file></filename> is omitted entirely, then the
1019 program behaves as a filter.</para>
1021 <para><command>hp2ps</command> is distributed in
1022 <filename>ghc/utils/hp2ps</filename> in a GHC source
1023 distribution. It was originally developed by Dave Wakeling as part
1024 of the HBC/LML heap profiler.</para>
1026 <para>The flags are:</para>
1031 <term><option>-d</option></term>
1033 <para>In order to make graphs more readable,
1034 <command>hp2ps</command> sorts the shaded bands for each
1035 identifier. The default sort ordering is for the bands with
1036 the largest area to be stacked on top of the smaller ones.
1037 The <option>-d</option> option causes rougher bands (those
1038 representing series of values with the largest standard
1039 deviations) to be stacked on top of smoother ones.</para>
1044 <term><option>-b</option></term>
1046 <para>Normally, <command>hp2ps</command> puts the title of
1047 the graph in a small box at the top of the page. However, if
1048 the JOB string is too long to fit in a small box (more than
1049 35 characters), then <command>hp2ps</command> will choose to
1050 use a big box instead. The <option>-b</option> option
1051 forces <command>hp2ps</command> to use a big box.</para>
1056 <term><option>-e<float>[in|mm|pt]</option></term>
1058 <para>Generate encapsulated PostScript suitable for
1059 inclusion in LaTeX documents. Usually, the PostScript graph
1060 is drawn in landscape mode in an area 9 inches wide by 6
1061 inches high, and <command>hp2ps</command> arranges for this
1062 area to be approximately centred on a sheet of a4 paper.
1063 This format is convenient of studying the graph in detail,
1064 but it is unsuitable for inclusion in LaTeX documents. The
1065 <option>-e</option> option causes the graph to be drawn in
1066 portrait mode, with float specifying the width in inches,
1067 millimetres or points (the default). The resulting
1068 PostScript file conforms to the Encapsulated PostScript
1069 (EPS) convention, and it can be included in a LaTeX document
1070 using Rokicki's dvi-to-PostScript converter
1071 <command>dvips</command>.</para>
1076 <term><option>-g</option></term>
1078 <para>Create output suitable for the <command>gs</command>
1079 PostScript previewer (or similar). In this case the graph is
1080 printed in portrait mode without scaling. The output is
1081 unsuitable for a laser printer.</para>
1086 <term><option>-l</option></term>
1088 <para>Normally a profile is limited to 20 bands with
1089 additional identifiers being grouped into an
1090 <literal>OTHER</literal> band. The <option>-l</option> flag
1091 removes this 20 band and limit, producing as many bands as
1092 necessary. No key is produced as it won't fit!. It is useful
1093 for creation time profiles with many bands.</para>
1098 <term><option>-m<int></option></term>
1100 <para>Normally a profile is limited to 20 bands with
1101 additional identifiers being grouped into an
1102 <literal>OTHER</literal> band. The <option>-m</option> flag
1103 specifies an alternative band limit (the maximum is
1106 <para><option>-m0</option> requests the band limit to be
1107 removed. As many bands as necessary are produced. However no
1108 key is produced as it won't fit! It is useful for displaying
1109 creation time profiles with many bands.</para>
1114 <term><option>-p</option></term>
1116 <para>Use previous parameters. By default, the PostScript
1117 graph is automatically scaled both horizontally and
1118 vertically so that it fills the page. However, when
1119 preparing a series of graphs for use in a presentation, it
1120 is often useful to draw a new graph using the same scale,
1121 shading and ordering as a previous one. The
1122 <option>-p</option> flag causes the graph to be drawn using
1123 the parameters determined by a previous run of
1124 <command>hp2ps</command> on <filename>file</filename>. These
1125 are extracted from <filename>file@.aux</filename>.</para>
1130 <term><option>-s</option></term>
1132 <para>Use a small box for the title.</para>
1137 <term><option>-t<float></option></term>
1139 <para>Normally trace elements which sum to a total of less
1140 than 1% of the profile are removed from the
1141 profile. The <option>-t</option> option allows this
1142 percentage to be modified (maximum 5%).</para>
1144 <para><option>-t0</option> requests no trace elements to be
1145 removed from the profile, ensuring that all the data will be
1151 <term><option>-c</option></term>
1153 <para>Generate colour output.</para>
1158 <term><option>-y</option></term>
1160 <para>Ignore marks.</para>
1165 <term><option>-?</option></term>
1167 <para>Print out usage information.</para>
1173 <sect2 id="manipulating-hp">
1174 <title>Manipulating the hp file</title>
1176 <para>(Notes kindly offered by Jan-Willhem Maessen.)</para>
1179 The <filename>FOO.hp</filename> file produced when you ask for the
1180 heap profile of a program <filename>FOO</filename> is a text file with a particularly
1181 simple structure. Here's a representative example, with much of the
1182 actual data omitted:
1185 DATE "Thu Dec 26 18:17 2002"
1186 SAMPLE_UNIT "seconds"
1197 BEGIN_SAMPLE 11695.47
1200 The first four lines (<literal>JOB</literal>, <literal>DATE</literal>, <literal>SAMPLE_UNIT</literal>, <literal>VALUE_UNIT</literal>) form a
1201 header. Each block of lines starting with <literal>BEGIN_SAMPLE</literal> and ending
1202 with <literal>END_SAMPLE</literal> forms a single sample (you can think of this as a
1203 vertical slice of your heap profile). The hp2ps utility should accept
1204 any input with a properly-formatted header followed by a series of
1210 <title>Zooming in on regions of your profile</title>
1213 You can look at particular regions of your profile simply by loading a
1214 copy of the <filename>.hp</filename> file into a text editor and deleting the unwanted
1215 samples. The resulting <filename>.hp</filename> file can be run through <command>hp2ps</command> and viewed
1221 <title>Viewing the heap profile of a running program</title>
1224 The <filename>.hp</filename> file is generated incrementally as your
1225 program runs. In principle, running <command>hp2ps</command> on the incomplete file
1226 should produce a snapshot of your program's heap usage. However, the
1227 last sample in the file may be incomplete, causing <command>hp2ps</command> to fail. If
1228 you are using a machine with UNIX utilities installed, it's not too
1229 hard to work around this problem (though the resulting command line
1230 looks rather Byzantine):
1232 head -`fgrep -n END_SAMPLE FOO.hp | tail -1 | cut -d : -f 1` FOO.hp \
1236 The command <command>fgrep -n END_SAMPLE FOO.hp</command> finds the
1237 end of every complete sample in <filename>FOO.hp</filename>, and labels each sample with
1238 its ending line number. We then select the line number of the last
1239 complete sample using <command>tail</command> and <command>cut</command>. This is used as a
1240 parameter to <command>head</command>; the result is as if we deleted the final
1241 incomplete sample from <filename>FOO.hp</filename>. This results in a properly-formatted
1242 .hp file which we feed directly to <command>hp2ps</command>.
1246 <title>Viewing a heap profile in real time</title>
1249 The <command>gv</command> and <command>ghostview</command> programs
1250 have a "watch file" option can be used to view an up-to-date heap
1251 profile of your program as it runs. Simply generate an incremental
1252 heap profile as described in the previous section. Run <command>gv</command> on your
1255 gv -watch -seascape FOO.ps
1257 If you forget the <literal>-watch</literal> flag you can still select
1258 "Watch file" from the "State" menu. Now each time you generate a new
1259 profile <filename>FOO.ps</filename> the view will update automatically.
1263 This can all be encapsulated in a little script:
1266 head -`fgrep -n END_SAMPLE FOO.hp | tail -1 | cut -d : -f 1` FOO.hp \
1268 gv -watch -seascape FOO.ps &
1270 sleep 10 # We generate a new profile every 10 seconds.
1271 head -`fgrep -n END_SAMPLE FOO.hp | tail -1 | cut -d : -f 1` FOO.hp \
1275 Occasionally <command>gv</command> will choke as it tries to read an incomplete copy of
1276 <filename>FOO.ps</filename> (because <command>hp2ps</command> is still running as an update
1277 occurs). A slightly more complicated script works around this
1278 problem, by using the fact that sending a SIGHUP to gv will cause it
1279 to re-read its input file:
1282 head -`fgrep -n END_SAMPLE FOO.hp | tail -1 | cut -d : -f 1` FOO.hp \
1288 head -`fgrep -n END_SAMPLE FOO.hp | tail -1 | cut -d : -f 1` FOO.hp \
1298 <title>Observing Code Coverage</title>
1299 <indexterm><primary>code coverage</primary></indexterm>
1300 <indexterm><primary>Haskell Program Coverage</primary></indexterm>
1301 <indexterm><primary>hpc</primary></indexterm>
1304 Code coverage tools allow a programmer to determine what parts of
1305 their code have been actually executed, and which parts have
1306 never actually been invoked. GHC has an option for generating
1307 instrumented code that records code coverage as part of the
1308 <ulink url="http://www.haskell.org/hpc">Haskell Program Coverage
1309 </ulink>(HPC) toolkit, which is included with GHC. HPC tools can
1310 be used to render the generated code coverage information into
1311 human understandable format. </para>
1314 Correctly instrumented code provides coverage information of two
1315 kinds: source coverage and boolean-control coverage. Source
1316 coverage is the extent to which every part of the program was
1317 used, measured at three different levels: declarations (both
1318 top-level and local), alternatives (among several equations or
1319 case branches) and expressions (at every level). Boolean
1320 coverage is the extent to which each of the values True and
1321 False is obtained in every syntactic boolean context (ie. guard,
1322 condition, qualifier). </para>
1325 HPC displays both kinds of information in two primary ways:
1326 textual reports with summary statistics (hpc report) and sources
1327 with color mark-up (hpc markup). For boolean coverage, there
1328 are four possible outcomes for each guard, condition or
1329 qualifier: both True and False values occur; only True; only
1330 False; never evaluated. In hpc-markup output, highlighting with
1331 a yellow background indicates a part of the program that was
1332 never evaluated; a green background indicates an always-True
1333 expression and a red background indicates an always-False one.
1336 <sect2><title>A small example: Reciprocation</title>
1339 For an example we have a program, called Recip.hs, which computes exact decimal
1340 representations of reciprocals, with recurring parts indicated in
1344 reciprocal :: Int -> (String, Int)
1345 reciprocal n | n > 1 = ('0' : '.' : digits, recur)
1347 "attempting to compute reciprocal of number <= 1"
1349 (digits, recur) = divide n 1 []
1350 divide :: Int -> Int -> [Int] -> (String, Int)
1351 divide n c cs | c `elem` cs = ([], position c cs)
1352 | r == 0 = (show q, 0)
1353 | r /= 0 = (show q ++ digits, recur)
1355 (q, r) = (c*10) `quotRem` n
1356 (digits, recur) = divide n r (c:cs)
1358 position :: Int -> [Int] -> Int
1359 position n (x:xs) | n==x = 1
1360 | otherwise = 1 + position n xs
1362 showRecip :: Int -> String
1364 "1/" ++ show n ++ " = " ++
1365 if r==0 then d else take p d ++ "(" ++ drop p d ++ ")"
1368 (d, r) = reciprocal n
1372 putStrLn (showRecip number)
1376 <para>The HPC instrumentation is enabled using the -fhpc flag.
1380 $ ghc -fhpc Recip.hs --make
1382 <para>HPC index (.mix) files are placed placed in .hpc subdirectory. These can be considered like
1383 the .hi files for HPC.
1390 <para>We can generate a textual summary of coverage:</para>
1393 80% expressions used (81/101)
1394 12% boolean coverage (1/8)
1395 14% guards (1/7), 3 always True,
1398 0% 'if' conditions (0/1), 1 always False
1399 100% qualifiers (0/0)
1400 55% alternatives used (5/9)
1401 100% local declarations used (9/9)
1402 100% top-level declarations used (5/5)
1404 <para>We can also generate a marked-up version of the source.</para>
1407 writing Recip.hs.html
1410 This generates one file per Haskell module, and 4 index files,
1411 hpc_index.html, hpc_index_alt.html, hpc_index_exp.html,
1416 <sect2><title>Options for instrumenting code for coverage</title>
1418 Turning on code coverage is easy, use the -fhpc flag.
1419 Instrumented and non-instrumented can be freely mixed.
1420 When compiling the Main module GHC automatically detects when there
1421 is an hpc compiled file, and adds the correct initialization code.
1426 <sect2><title>The hpc toolkit</title>
1429 The hpc toolkit uses a cvs/svn/darcs-like interface, where a
1430 single binary contains many function units.</para>
1433 Usage: hpc COMMAND ...
1436 help Display help for hpc or a single command
1438 report Output textual report about program coverage
1439 markup Markup Haskell source with program coverage
1440 Processing Coverage files:
1441 sum Sum multiple .tix files in a single .tix file
1442 combine Combine two .tix files in a single .tix file
1443 map Map a function over a single .tix file
1445 overlay Generate a .tix file from an overlay file
1446 draft Generate draft overlay that provides 100% coverage
1448 show Show .tix file in readable, verbose format
1449 version Display version for hpc
1452 <para>In general, these options act on .tix file after an
1453 instrumented binary has generated it, which hpc acting as a
1454 conduit between the raw .tix file, and the more detailed reports
1459 The hpc tool assumes you are in the top-level directory of
1460 the location where you built your application, and the .tix
1461 file is in the same top-level directory. You can use the
1462 flag --srcdir to use hpc for any other directory, and use
1463 --srcdir multiple times to analyse programs compiled from
1464 difference locations, as is typical for packages.
1468 We now explain in more details the major modes of hpc.
1471 <sect3><title>hpc report</title>
1472 <para>hpc report gives a textual report of coverage. By default,
1473 all modules and packages are considered in generating report,
1474 unless include or exclude are used. The report is a summary
1475 unless the --per-module flag is used. The --xml-output option
1476 allows for tools to use hpc to glean coverage.
1480 Usage: hpc report [OPTION] .. <TIX_FILE> [<MODULE> [<MODULE> ..]]
1484 --per-module show module level detail
1485 --decl-list show unused decls
1486 --exclude=[PACKAGE:][MODULE] exclude MODULE and/or PACKAGE
1487 --include=[PACKAGE:][MODULE] include MODULE and/or PACKAGE
1488 --srcdir=DIR path to source directory of .hs files
1489 multi-use of srcdir possible
1490 --hpcdir=DIR sub-directory that contains .mix files
1491 default .hpc [rarely used]
1492 --xml-output show output in XML
1495 <sect3><title>hpc markup</title>
1496 <para>hpc markup marks up source files into colored html.
1500 Usage: hpc markup [OPTION] .. <TIX_FILE> [<MODULE> [<MODULE> ..]]
1504 --exclude=[PACKAGE:][MODULE] exclude MODULE and/or PACKAGE
1505 --include=[PACKAGE:][MODULE] include MODULE and/or PACKAGE
1506 --srcdir=DIR path to source directory of .hs files
1507 multi-use of srcdir possible
1508 --hpcdir=DIR sub-directory that contains .mix files
1509 default .hpc [rarely used]
1510 --fun-entry-count show top-level function entry counts
1511 --highlight-covered highlight covered code, rather that code gaps
1512 --destdir=DIR path to write output to
1516 <sect3><title>hpc sum</title>
1517 <para>hpc sum adds together any number of .tix files into a single
1518 .tix file. hpc sum does not change the original .tix file; it generates a new .tix file.
1522 Usage: hpc sum [OPTION] .. <TIX_FILE> [<TIX_FILE> [<TIX_FILE> ..]]
1523 Sum multiple .tix files in a single .tix file
1527 --exclude=[PACKAGE:][MODULE] exclude MODULE and/or PACKAGE
1528 --include=[PACKAGE:][MODULE] include MODULE and/or PACKAGE
1529 --output=FILE output FILE
1530 --union use the union of the module namespace (default is intersection)
1533 <sect3><title>hpc combine</title>
1534 <para>hpc combine is the swiss army knife of hpc. It can be
1535 used to take the difference between .tix files, to subtract one
1536 .tix file from another, or to add two .tix files. hpc combine does not
1537 change the original .tix file; it generates a new .tix file.
1541 Usage: hpc combine [OPTION] .. <TIX_FILE> <TIX_FILE>
1542 Combine two .tix files in a single .tix file
1546 --exclude=[PACKAGE:][MODULE] exclude MODULE and/or PACKAGE
1547 --include=[PACKAGE:][MODULE] include MODULE and/or PACKAGE
1548 --output=FILE output FILE
1549 --function=FUNCTION combine .tix files with join function, default = ADD
1550 FUNCTION = ADD | DIFF | SUB
1551 --union use the union of the module namespace (default is intersection)
1554 <sect3><title>hpc map</title>
1555 <para>hpc map inverts or zeros a .tix file. hpc map does not
1556 change the original .tix file; it generates a new .tix file.
1560 Usage: hpc map [OPTION] .. <TIX_FILE>
1561 Map a function over a single .tix file
1565 --exclude=[PACKAGE:][MODULE] exclude MODULE and/or PACKAGE
1566 --include=[PACKAGE:][MODULE] include MODULE and/or PACKAGE
1567 --output=FILE output FILE
1568 --function=FUNCTION apply function to .tix files, default = ID
1569 FUNCTION = ID | INV | ZERO
1570 --union use the union of the module namespace (default is intersection)
1573 <sect3><title>hpc overlay and hpc draft</title>
1575 Overlays are an experimental feature of HPC, a textual description
1576 of coverage. hpc draft is used to generate a draft overlay from a .tix file,
1577 and hpc overlay generates a .tix files from an overlay.
1581 Usage: hpc overlay [OPTION] .. <OVERLAY_FILE> [<OVERLAY_FILE> [...]]
1585 --srcdir=DIR path to source directory of .hs files
1586 multi-use of srcdir possible
1587 --hpcdir=DIR sub-directory that contains .mix files
1588 default .hpc [rarely used]
1589 --output=FILE output FILE
1591 Usage: hpc draft [OPTION] .. <TIX_FILE>
1595 --exclude=[PACKAGE:][MODULE] exclude MODULE and/or PACKAGE
1596 --include=[PACKAGE:][MODULE] include MODULE and/or PACKAGE
1597 --srcdir=DIR path to source directory of .hs files
1598 multi-use of srcdir possible
1599 --hpcdir=DIR sub-directory that contains .mix files
1600 default .hpc [rarely used]
1601 --output=FILE output FILE
1605 <sect2><title>Caveats and Shortcomings of Haskell Program Coverage</title>
1607 HPC does not attempt to lock the .tix file, so multiple concurrently running
1608 binaries in the same directory will exhibit a race condition. There is no way
1609 to change the name of the .tix file generated, apart from renaming the binary.
1610 HPC does not work with GHCi.
1615 <sect1 id="ticky-ticky">
1616 <title>Using “ticky-ticky” profiling (for implementors)</title>
1617 <indexterm><primary>ticky-ticky profiling</primary></indexterm>
1619 <para>(ToDo: document properly.)</para>
1621 <para>It is possible to compile Haskell programs so that
1622 they will count lots and lots of interesting things, e.g., number
1623 of updates, number of data constructors entered, etc., etc. We
1624 call this “ticky-ticky”
1625 profiling,<indexterm><primary>ticky-ticky
1626 profiling</primary></indexterm> <indexterm><primary>profiling,
1627 ticky-ticky</primary></indexterm> because that's the sound a CPU
1628 makes when it is running up all those counters
1629 (<emphasis>slowly</emphasis>).</para>
1631 <para>Ticky-ticky profiling is mainly intended for implementors;
1632 it is quite separate from the main “cost-centre”
1633 profiling system, intended for all users everywhere.</para>
1636 You don't need to build GHC, the libraries, or the RTS a special
1637 way in order to use ticky-ticky profiling. You can decide on a
1638 module-by-module basis which parts of a program have the
1639 counters compiled in, using the
1640 compile-time <option>-ticky</option> option. Those modules that
1641 were not compiled with <option>-ticky</option> won't contribute
1642 to the ticky-ticky profiling results, and that will normally
1643 include all the pre-compiled packages that your program links
1648 To get your compiled program to spit out the ticky-ticky
1654 Link the program with <option>-debug</option>
1655 (<option>-ticky</option> is a synonym
1656 for <option>-debug</option> at link-time). This links in
1657 the debug version of the RTS, which includes the code for
1658 aggregating and reporting the results of ticky-ticky
1664 Run the program with the <option>-r</option> RTS
1665 option<indexterm><primary>-r RTS option</primary></indexterm>.
1666 See <xref linkend="runtime-control"/>.
1673 Here is a sample ticky-ticky statistics file, generated by
1675 <command>foo +RTS -rfoo.ticky</command>.
1679 foo +RTS -rfoo.ticky
1681 ALLOCATIONS: 3964631 (11330900 words total: 3999476 admin, 6098829 goods, 1232595 slop)
1682 total words: 2 3 4 5 6+
1683 69647 ( 1.8%) function values 50.0 50.0 0.0 0.0 0.0
1684 2382937 ( 60.1%) thunks 0.0 83.9 16.1 0.0 0.0
1685 1477218 ( 37.3%) data values 66.8 33.2 0.0 0.0 0.0
1686 0 ( 0.0%) big tuples
1687 2 ( 0.0%) black holes 0.0 100.0 0.0 0.0 0.0
1688 0 ( 0.0%) prim things
1689 34825 ( 0.9%) partial applications 0.0 0.0 0.0 100.0 0.0
1690 2 ( 0.0%) thread state objects 0.0 0.0 0.0 0.0 100.0
1692 Total storage-manager allocations: 3647137 (11882004 words)
1693 [551104 words lost to speculative heap-checks]
1697 ENTERS: 9400092 of which 2005772 (21.3%) direct to the entry code
1698 [the rest indirected via Node's info ptr]
1699 1860318 ( 19.8%) thunks
1700 3733184 ( 39.7%) data values
1701 3149544 ( 33.5%) function values
1702 [of which 1999880 (63.5%) bypassed arg-satisfaction chk]
1703 348140 ( 3.7%) partial applications
1704 308906 ( 3.3%) normal indirections
1705 0 ( 0.0%) permanent indirections
1708 2137257 ( 36.4%) from entering a new constructor
1709 [the rest from entering an existing constructor]
1710 2349219 ( 40.0%) vectored [the rest unvectored]
1712 RET_NEW: 2137257: 32.5% 46.2% 21.3% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0%
1713 RET_OLD: 3733184: 2.8% 67.9% 29.3% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0%
1714 RET_UNBOXED_TUP: 2: 0.0% 0.0%100.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0%
1716 RET_VEC_RETURN : 2349219: 0.0% 0.0%100.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0%
1718 UPDATE FRAMES: 2241725 (0 omitted from thunks)
1722 0 ( 0.0%) data values
1723 34827 ( 1.6%) partial applications
1724 [2 in place, 34825 allocated new space]
1725 2206898 ( 98.4%) updates to existing heap objects (46 by squeezing)
1726 UPD_CON_IN_NEW: 0: 0 0 0 0 0 0 0 0 0
1727 UPD_PAP_IN_NEW: 34825: 0 0 0 34825 0 0 0 0 0
1729 NEW GEN UPDATES: 2274700 ( 99.9%)
1731 OLD GEN UPDATES: 1852 ( 0.1%)
1733 Total bytes copied during GC: 190096
1735 **************************************************
1736 3647137 ALLOC_HEAP_ctr
1737 11882004 ALLOC_HEAP_tot
1742 34831 ALLOC_FUN_hst_0
1743 34816 ALLOC_FUN_hst_1
1747 2382937 ALLOC_UP_THK_ctr
1750 0 E!NT_PERM_IND_ctr requires +RTS -Z
1751 [... lots more info omitted ...]
1752 0 GC_SEL_ABANDONED_ctr
1755 0 GC_FAILED_PROMOTION_ctr
1756 47524 GC_WORDS_COPIED_ctr
1759 <para>The formatting of the information above the row of asterisks
1760 is subject to change, but hopefully provides a useful
1761 human-readable summary. Below the asterisks <emphasis>all
1762 counters</emphasis> maintained by the ticky-ticky system are
1763 dumped, in a format intended to be machine-readable: zero or more
1764 spaces, an integer, a space, the counter name, and a newline.</para>
1766 <para>In fact, not <emphasis>all</emphasis> counters are
1767 necessarily dumped; compile- or run-time flags can render certain
1768 counters invalid. In this case, either the counter will simply
1769 not appear, or it will appear with a modified counter name,
1770 possibly along with an explanation for the omission (notice
1771 <literal>ENT_PERM_IND_ctr</literal> appears
1772 with an inserted <literal>!</literal> above). Software analysing
1773 this output should always check that it has the counters it
1774 expects. Also, beware: some of the counters can have
1775 <emphasis>large</emphasis> values!</para>
1782 ;;; Local Variables: ***
1784 ;;; sgml-parent-document: ("users_guide.xml" "book" "chapter") ***