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.</para>
37 <indexterm><primary><option>-p</option></primary><secondary>RTS
38 option</secondary></indexterm>
42 <para> Examine the generated profiling information, using one of
43 GHC's profiling tools. The tool to use will depend on the kind
44 of profiling information generated.</para>
49 <sect1 id="cost-centres">
50 <title>Cost centres and cost-centre stacks</title>
52 <para>GHC's profiling system assigns <firstterm>costs</firstterm>
53 to <firstterm>cost centres</firstterm>. A cost is simply the time
54 or space required to evaluate an expression. Cost centres are
55 program annotations around expressions; all costs incurred by the
56 annotated expression are assigned to the enclosing cost centre.
57 Furthermore, GHC will remember the stack of enclosing cost centres
58 for any given expression at run-time and generate a call-graph of
59 cost attributions.</para>
61 <para>Let's take a look at an example:</para>
64 main = print (nfib 25)
65 nfib n = if n < 2 then 1 else nfib (n-1) + nfib (n-2)
68 <para>Compile and run this program as follows:</para>
71 $ ghc -prof -auto-all -o Main Main.hs
77 <para>When a GHC-compiled program is run with the
78 <option>-p</option> RTS option, it generates a file called
79 <filename><prog>.prof</filename>. In this case, the file
80 will contain something like this:</para>
83 Fri May 12 14:06 2000 Time and Allocation Profiling Report (Final)
87 total time = 0.14 secs (7 ticks @ 20 ms)
88 total alloc = 8,741,204 bytes (excludes profiling overheads)
90 COST CENTRE MODULE %time %alloc
96 COST CENTRE MODULE entries %time %alloc %time %alloc
98 MAIN MAIN 0 0.0 0.0 100.0 100.0
99 main Main 0 0.0 0.0 0.0 0.0
100 CAF PrelHandle 3 0.0 0.0 0.0 0.0
101 CAF PrelAddr 1 0.0 0.0 0.0 0.0
102 CAF Main 6 0.0 0.0 100.0 100.0
103 main Main 1 0.0 0.0 100.0 100.0
104 nfib Main 242785 100.0 100.0 100.0 100.0
108 <para>The first part of the file gives the program name and
109 options, and the total time and total memory allocation measured
110 during the run of the program (note that the total memory
111 allocation figure isn't the same as the amount of
112 <emphasis>live</emphasis> memory needed by the program at any one
113 time; the latter can be determined using heap profiling, which we
114 will describe shortly).</para>
116 <para>The second part of the file is a break-down by cost centre
117 of the most costly functions in the program. In this case, there
118 was only one significant function in the program, namely
119 <function>nfib</function>, and it was responsible for 100%
120 of both the time and allocation costs of the program.</para>
122 <para>The third and final section of the file gives a profile
123 break-down by cost-centre stack. This is roughly a call-graph
124 profile of the program. In the example above, it is clear that
125 the costly call to <function>nfib</function> came from
126 <function>main</function>.</para>
128 <para>The time and allocation incurred by a given part of the
129 program is displayed in two ways: “individual”, which
130 are the costs incurred by the code covered by this cost centre
131 stack alone, and “inherited”, which includes the costs
132 incurred by all the children of this node.</para>
134 <para>The usefulness of cost-centre stacks is better demonstrated
135 by modifying the example slightly:</para>
138 main = print (f 25 + g 25)
140 g n = nfib (n `div` 2)
141 nfib n = if n < 2 then 1 else nfib (n-1) + nfib (n-2)
144 <para>Compile and run this program as before, and take a look at
145 the new profiling results:</para>
148 COST CENTRE MODULE scc %time %alloc %time %alloc
150 MAIN MAIN 0 0.0 0.0 100.0 100.0
151 main Main 0 0.0 0.0 0.0 0.0
152 CAF PrelHandle 3 0.0 0.0 0.0 0.0
153 CAF PrelAddr 1 0.0 0.0 0.0 0.0
154 CAF Main 9 0.0 0.0 100.0 100.0
155 main Main 1 0.0 0.0 100.0 100.0
156 g Main 1 0.0 0.0 0.0 0.2
157 nfib Main 465 0.0 0.2 0.0 0.2
158 f Main 1 0.0 0.0 100.0 99.8
159 nfib Main 242785 100.0 99.8 100.0 99.8
162 <para>Now although we had two calls to <function>nfib</function>
163 in the program, it is immediately clear that it was the call from
164 <function>f</function> which took all the time.</para>
166 <para>The actual meaning of the various columns in the output is:</para>
172 <para>The number of times this particular point in the call
173 graph was entered.</para>
178 <term>individual %time</term>
180 <para>The percentage of the total run time of the program
181 spent at this point in the call graph.</para>
186 <term>individual %alloc</term>
188 <para>The percentage of the total memory allocations
189 (excluding profiling overheads) of the program made by this
195 <term>inherited %time</term>
197 <para>The percentage of the total run time of the program
198 spent below this point in the call graph.</para>
203 <term>inherited %alloc</term>
205 <para>The percentage of the total memory allocations
206 (excluding profiling overheads) of the program made by this
207 call and all of its sub-calls.</para>
212 <para>In addition you can use the <option>-P</option> RTS option
213 <indexterm><primary><option>-P</option></primary></indexterm> to
214 get the following additional information:</para>
218 <term><literal>ticks</literal></term>
220 <para>The raw number of time “ticks” which were
221 attributed to this cost-centre; from this, we get the
222 <literal>%time</literal> figure mentioned
228 <term><literal>bytes</literal></term>
230 <para>Number of bytes allocated in the heap while in this
231 cost-centre; again, this is the raw number from which we get
232 the <literal>%alloc</literal> figure mentioned
238 <para>What about recursive functions, and mutually recursive
239 groups of functions? Where are the costs attributed? Well,
240 although GHC does keep information about which groups of functions
241 called each other recursively, this information isn't displayed in
242 the basic time and allocation profile, instead the call-graph is
243 flattened into a tree.</para>
245 <sect2><title>Inserting cost centres by hand</title>
247 <para>Cost centres are just program annotations. When you say
248 <option>-auto-all</option> to the compiler, it automatically
249 inserts a cost centre annotation around every top-level function
250 in your program, but you are entirely free to add the cost
251 centre annotations yourself.</para>
253 <para>The syntax of a cost centre annotation is</para>
256 {-# SCC "name" #-} <expression>
259 <para>where <literal>"name"</literal> is an arbitrary string,
260 that will become the name of your cost centre as it appears
261 in the profiling output, and
262 <literal><expression></literal> is any Haskell
263 expression. An <literal>SCC</literal> annotation extends as
264 far to the right as possible when parsing. (SCC stands for "Set
265 Cost Centre").</para>
269 <sect2 id="prof-rules">
270 <title>Rules for attributing costs</title>
272 <para>The cost of evaluating any expression in your program is
273 attributed to a cost-centre stack using the following rules:</para>
277 <para>If the expression is part of the
278 <firstterm>one-off</firstterm> costs of evaluating the
279 enclosing top-level definition, then costs are attributed to
280 the stack of lexically enclosing <literal>SCC</literal>
281 annotations on top of the special <literal>CAF</literal>
286 <para>Otherwise, costs are attributed to the stack of
287 lexically-enclosing <literal>SCC</literal> annotations,
288 appended to the cost-centre stack in effect at the
289 <firstterm>call site</firstterm> of the current top-level
290 definition<footnote> <para>The call-site is just the place
291 in the source code which mentions the particular function or
292 variable.</para></footnote>. Notice that this is a recursive
297 <para>Time spent in foreign code (see <xref linkend="ffi"/>)
298 is always attributed to the cost centre in force at the
299 Haskell call-site of the foreign function.</para>
303 <para>What do we mean by one-off costs? Well, Haskell is a lazy
304 language, and certain expressions are only ever evaluated once.
305 For example, if we write:</para>
311 <para>then <varname>x</varname> will only be evaluated once (if
312 at all), and subsequent demands for <varname>x</varname> will
313 immediately get to see the cached result. The definition
314 <varname>x</varname> is called a CAF (Constant Applicative
315 Form), because it has no arguments.</para>
317 <para>For the purposes of profiling, we say that the expression
318 <literal>nfib 25</literal> belongs to the one-off costs of
319 evaluating <varname>x</varname>.</para>
321 <para>Since one-off costs aren't strictly speaking part of the
322 call-graph of the program, they are attributed to a special
323 top-level cost centre, <literal>CAF</literal>. There may be one
324 <literal>CAF</literal> cost centre for each module (the
325 default), or one for each top-level definition with any one-off
326 costs (this behaviour can be selected by giving GHC the
327 <option>-caf-all</option> flag).</para>
329 <indexterm><primary><literal>-caf-all</literal></primary>
332 <para>If you think you have a weird profile, or the call-graph
333 doesn't look like you expect it to, feel free to send it (and
334 your program) to us at
335 <email>glasgow-haskell-bugs@haskell.org</email>.</para>
339 <sect1 id="prof-compiler-options">
340 <title>Compiler options for profiling</title>
342 <indexterm><primary>profiling</primary><secondary>options</secondary></indexterm>
343 <indexterm><primary>options</primary><secondary>for profiling</secondary></indexterm>
348 <option>-prof</option>:
349 <indexterm><primary><option>-prof</option></primary></indexterm>
352 <para> To make use of the profiling system
353 <emphasis>all</emphasis> modules must be compiled and linked
354 with the <option>-prof</option> option. Any
355 <literal>SCC</literal> annotations you've put in your source
356 will spring to life.</para>
358 <para> Without a <option>-prof</option> option, your
359 <literal>SCC</literal>s are ignored; so you can compile
360 <literal>SCC</literal>-laden code without changing
366 <para>There are a few other profiling-related compilation options.
367 Use them <emphasis>in addition to</emphasis>
368 <option>-prof</option>. These do not have to be used consistently
369 for all modules in a program.</para>
374 <option>-auto</option>:
375 <indexterm><primary><option>-auto</option></primary></indexterm>
376 <indexterm><primary>cost centres</primary><secondary>automatically inserting</secondary></indexterm>
379 <para> GHC will automatically add
380 <function>_scc_</function> constructs for all
381 top-level, exported functions.</para>
387 <option>-auto-all</option>:
388 <indexterm><primary><option>-auto-all</option></primary></indexterm>
391 <para> <emphasis>All</emphasis> top-level functions,
392 exported or not, will be automatically
393 <function>_scc_</function>'d.</para>
399 <option>-caf-all</option>:
400 <indexterm><primary><option>-caf-all</option></primary></indexterm>
403 <para> The costs of all CAFs in a module are usually
404 attributed to one “big” CAF cost-centre. With
405 this option, all CAFs get their own cost-centre. An
406 “if all else fails” option…</para>
412 <option>-ignore-scc</option>:
413 <indexterm><primary><option>-ignore-scc</option></primary></indexterm>
416 <para>Ignore any <function>_scc_</function>
417 constructs, so a module which already has
418 <function>_scc_</function>s can be compiled
419 for profiling with the annotations ignored.</para>
427 <sect1 id="prof-time-options">
428 <title>Time and allocation profiling</title>
430 <para>To generate a time and allocation profile, give one of the
431 following RTS options to the compiled program when you run it (RTS
432 options should be enclosed between <literal>+RTS...-RTS</literal>
438 <option>-p</option> or <option>-P</option>:
439 <indexterm><primary><option>-p</option></primary></indexterm>
440 <indexterm><primary><option>-P</option></primary></indexterm>
441 <indexterm><primary>time profile</primary></indexterm>
444 <para>The <option>-p</option> option produces a standard
445 <emphasis>time profile</emphasis> report. It is written
447 <filename><replaceable>program</replaceable>.prof</filename>.</para>
449 <para>The <option>-P</option> option produces a more
450 detailed report containing the actual time and allocation
451 data as well. (Not used much.)</para>
458 <indexterm><primary><option>-xc</option></primary><secondary>RTS option</secondary></indexterm>
461 <para>This option makes use of the extra information
462 maintained by the cost-centre-stack profiler to provide
463 useful information about the location of runtime errors.
464 See <xref linkend="rts-options-debugging"/>.</para>
472 <sect1 id="prof-heap">
473 <title>Profiling memory usage</title>
475 <para>In addition to profiling the time and allocation behaviour
476 of your program, you can also generate a graph of its memory usage
477 over time. This is useful for detecting the causes of
478 <firstterm>space leaks</firstterm>, when your program holds on to
479 more memory at run-time that it needs to. Space leaks lead to
480 longer run-times due to heavy garbage collector activity, and may
481 even cause the program to run out of memory altogether.</para>
483 <para>To generate a heap profile from your program:</para>
487 <para>Compile the program for profiling (<xref
488 linkend="prof-compiler-options"/>).</para>
491 <para>Run it with one of the heap profiling options described
492 below (eg. <option>-hc</option> for a basic producer profile).
493 This generates the file
494 <filename><replaceable>prog</replaceable>.hp</filename>.</para>
497 <para>Run <command>hp2ps</command> to produce a Postscript
499 <filename><replaceable>prog</replaceable>.ps</filename>. The
500 <command>hp2ps</command> utility is described in detail in
501 <xref linkend="hp2ps"/>.</para>
504 <para>Display the heap profile using a postscript viewer such
505 as <application>Ghostview</application>, or print it out on a
506 Postscript-capable printer.</para>
510 <sect2 id="rts-options-heap-prof">
511 <title>RTS options for heap profiling</title>
513 <para>There are several different kinds of heap profile that can
514 be generated. All the different profile types yield a graph of
515 live heap against time, but they differ in how the live heap is
516 broken down into bands. The following RTS options select which
517 break-down to use:</para>
523 <indexterm><primary><option>-hc</option></primary><secondary>RTS option</secondary></indexterm>
526 <para>Breaks down the graph by the cost-centre stack which
527 produced the data.</para>
534 <indexterm><primary><option>-hm</option></primary><secondary>RTS option</secondary></indexterm>
537 <para>Break down the live heap by the module containing
538 the code which produced the data.</para>
545 <indexterm><primary><option>-hd</option></primary><secondary>RTS option</secondary></indexterm>
548 <para>Breaks down the graph by <firstterm>closure
549 description</firstterm>. For actual data, the description
550 is just the constructor name, for other closures it is a
551 compiler-generated string identifying the closure.</para>
558 <indexterm><primary><option>-hy</option></primary><secondary>RTS option</secondary></indexterm>
561 <para>Breaks down the graph by
562 <firstterm>type</firstterm>. For closures which have
563 function type or unknown/polymorphic type, the string will
564 represent an approximation to the actual type.</para>
571 <indexterm><primary><option>-hr</option></primary><secondary>RTS option</secondary></indexterm>
574 <para>Break down the graph by <firstterm>retainer
575 set</firstterm>. Retainer profiling is described in more
576 detail below (<xref linkend="retainer-prof"/>).</para>
583 <indexterm><primary><option>-hb</option></primary><secondary>RTS option</secondary></indexterm>
586 <para>Break down the graph by
587 <firstterm>biography</firstterm>. Biographical profiling
588 is described in more detail below (<xref
589 linkend="biography-prof"/>).</para>
594 <para>In addition, the profile can be restricted to heap data
595 which satisfies certain criteria - for example, you might want
596 to display a profile by type but only for data produced by a
597 certain module, or a profile by retainer for a certain type of
598 data. Restrictions are specified as follows:</para>
603 <option>-hc</option><replaceable>name</replaceable>,...
604 <indexterm><primary><option>-hc</option></primary><secondary>RTS option</secondary></indexterm>
607 <para>Restrict the profile to closures produced by
608 cost-centre stacks with one of the specified cost centres
615 <option>-hC</option><replaceable>name</replaceable>,...
616 <indexterm><primary><option>-hC</option></primary><secondary>RTS option</secondary></indexterm>
619 <para>Restrict the profile to closures produced by
620 cost-centre stacks with one of the specified cost centres
621 anywhere in the stack.</para>
627 <option>-hm</option><replaceable>module</replaceable>,...
628 <indexterm><primary><option>-hm</option></primary><secondary>RTS option</secondary></indexterm>
631 <para>Restrict the profile to closures produced by the
632 specified modules.</para>
638 <option>-hd</option><replaceable>desc</replaceable>,...
639 <indexterm><primary><option>-hd</option></primary><secondary>RTS option</secondary></indexterm>
642 <para>Restrict the profile to closures with the specified
643 description strings.</para>
649 <option>-hy</option><replaceable>type</replaceable>,...
650 <indexterm><primary><option>-hy</option></primary><secondary>RTS option</secondary></indexterm>
653 <para>Restrict the profile to closures with the specified
660 <option>-hr</option><replaceable>cc</replaceable>,...
661 <indexterm><primary><option>-hr</option></primary><secondary>RTS option</secondary></indexterm>
664 <para>Restrict the profile to closures with retainer sets
665 containing cost-centre stacks with one of the specified
666 cost centres at the top.</para>
672 <option>-hb</option><replaceable>bio</replaceable>,...
673 <indexterm><primary><option>-hb</option></primary><secondary>RTS option</secondary></indexterm>
676 <para>Restrict the profile to closures with one of the
677 specified biographies, where
678 <replaceable>bio</replaceable> is one of
679 <literal>lag</literal>, <literal>drag</literal>,
680 <literal>void</literal>, or <literal>use</literal>.</para>
685 <para>For example, the following options will generate a
686 retainer profile restricted to <literal>Branch</literal> and
687 <literal>Leaf</literal> constructors:</para>
690 <replaceable>prog</replaceable> +RTS -hr -hdBranch,Leaf
693 <para>There can only be one "break-down" option
694 (eg. <option>-hr</option> in the example above), but there is no
695 limit on the number of further restrictions that may be applied.
696 All the options may be combined, with one exception: GHC doesn't
697 currently support mixing the <option>-hr</option> and
698 <option>-hb</option> options.</para>
700 <para>There are three more options which relate to heap
706 <option>-i<replaceable>secs</replaceable></option>:
707 <indexterm><primary><option>-i</option></primary></indexterm>
710 <para>Set the profiling (sampling) interval to
711 <replaceable>secs</replaceable> seconds (the default is
712 0.1 second). Fractions are allowed: for example
713 <option>-i0.2</option> will get 5 samples per second.
714 This only affects heap profiling; time profiles are always
715 sampled on a 1/50 second frequency.</para>
722 <indexterm><primary><option>-xt</option></primary><secondary>RTS option</secondary></indexterm>
725 <para>Include the memory occupied by threads in a heap
726 profile. Each thread takes up a small area for its thread
727 state in addition to the space allocated for its stack
728 (stacks normally start small and then grow as
731 <para>This includes the main thread, so using
732 <option>-xt</option> is a good way to see how much stack
733 space the program is using.</para>
735 <para>Memory occupied by threads and their stacks is
736 labelled as “TSO” when displaying the profile
737 by closure description or type description.</para>
743 <option>-L<replaceable>num</replaceable></option>
744 <indexterm><primary><option>-L</option></primary><secondary>RTS option</secondary></indexterm>
748 Sets the maximum length of a cost-centre stack name in a
749 heap profile. Defaults to 25.
757 <sect2 id="retainer-prof">
758 <title>Retainer Profiling</title>
760 <para>Retainer profiling is designed to help answer questions
761 like <quote>why is this data being retained?</quote>. We start
762 by defining what we mean by a retainer:</para>
765 <para>A retainer is either the system stack, or an unevaluated
766 closure (thunk).</para>
769 <para>In particular, constructors are <emphasis>not</emphasis>
772 <para>An object B retains object A if (i) B is a retainer object and
773 (ii) object A can be reached by recursively following pointers
774 starting from object B, but not meeting any other retainer
775 objects on the way. Each live object is retained by one or more
776 retainer objects, collectively called its retainer set, or its
777 <firstterm>retainer set</firstterm>, or its
778 <firstterm>retainers</firstterm>.</para>
780 <para>When retainer profiling is requested by giving the program
781 the <option>-hr</option> option, a graph is generated which is
782 broken down by retainer set. A retainer set is displayed as a
783 set of cost-centre stacks; because this is usually too large to
784 fit on the profile graph, each retainer set is numbered and
785 shown abbreviated on the graph along with its number, and the
786 full list of retainer sets is dumped into the file
787 <filename><replaceable>prog</replaceable>.prof</filename>.</para>
789 <para>Retainer profiling requires multiple passes over the live
790 heap in order to discover the full retainer set for each
791 object, which can be quite slow. So we set a limit on the
792 maximum size of a retainer set, where all retainer sets larger
793 than the maximum retainer set size are replaced by the special
794 set <literal>MANY</literal>. The maximum set size defaults to 8
795 and can be altered with the <option>-R</option> RTS
800 <term><option>-R</option><replaceable>size</replaceable></term>
802 <para>Restrict the number of elements in a retainer set to
803 <replaceable>size</replaceable> (default 8).</para>
809 <title>Hints for using retainer profiling</title>
811 <para>The definition of retainers is designed to reflect a
812 common cause of space leaks: a large structure is retained by
813 an unevaluated computation, and will be released once the
814 computation is forced. A good example is looking up a value in
815 a finite map, where unless the lookup is forced in a timely
816 manner the unevaluated lookup will cause the whole mapping to
817 be retained. These kind of space leaks can often be
818 eliminated by forcing the relevant computations to be
819 performed eagerly, using <literal>seq</literal> or strictness
820 annotations on data constructor fields.</para>
822 <para>Often a particular data structure is being retained by a
823 chain of unevaluated closures, only the nearest of which will
824 be reported by retainer profiling - for example A retains B, B
825 retains C, and C retains a large structure. There might be a
826 large number of Bs but only a single A, so A is really the one
827 we're interested in eliminating. However, retainer profiling
828 will in this case report B as the retainer of the large
829 structure. To move further up the chain of retainers, we can
830 ask for another retainer profile but this time restrict the
831 profile to B objects, so we get a profile of the retainers of
835 <replaceable>prog</replaceable> +RTS -hr -hcB
838 <para>This trick isn't foolproof, because there might be other
839 B closures in the heap which aren't the retainers we are
840 interested in, but we've found this to be a useful technique
841 in most cases.</para>
845 <sect2 id="biography-prof">
846 <title>Biographical Profiling</title>
848 <para>A typical heap object may be in one of the following four
849 states at each point in its lifetime:</para>
853 <para>The <firstterm>lag</firstterm> stage, which is the
854 time between creation and the first use of the
858 <para>the <firstterm>use</firstterm> stage, which lasts from
859 the first use until the last use of the object, and</para>
862 <para>The <firstterm>drag</firstterm> stage, which lasts
863 from the final use until the last reference to the object
867 <para>An object which is never used is said to be in the
868 <firstterm>void</firstterm> state for its whole
873 <para>A biographical heap profile displays the portion of the
874 live heap in each of the four states listed above. Usually the
875 most interesting states are the void and drag states: live heap
876 in these states is more likely to be wasted space than heap in
877 the lag or use states.</para>
879 <para>It is also possible to break down the heap in one or more
880 of these states by a different criteria, by restricting a
881 profile by biography. For example, to show the portion of the
882 heap in the drag or void state by producer: </para>
885 <replaceable>prog</replaceable> +RTS -hc -hbdrag,void
888 <para>Once you know the producer or the type of the heap in the
889 drag or void states, the next step is usually to find the
893 <replaceable>prog</replaceable> +RTS -hr -hc<replaceable>cc</replaceable>...
896 <para>NOTE: this two stage process is required because GHC
897 cannot currently profile using both biographical and retainer
898 information simultaneously.</para>
901 <sect2 id="mem-residency">
902 <title>Actual memory residency</title>
904 <para>How does the heap residency reported by the heap profiler relate to
905 the actual memory residency of your program when you run it? You might
906 see a large discrepancy between the residency reported by the heap
907 profiler, and the residency reported by tools on your system
908 (eg. <literal>ps</literal> or <literal>top</literal> on Unix, or the
909 Task Manager on Windows). There are several reasons for this:</para>
913 <para>There is an overhead of profiling itself, which is subtracted
914 from the residency figures by the profiler. This overhead goes
915 away when compiling without profiling support, of course. The
916 space overhead is currently 2 extra
917 words per heap object, which probably results in
918 about a 30% overhead.</para>
922 <para>Garbage collection requires more memory than the actual
923 residency. The factor depends on the kind of garbage collection
924 algorithm in use: a major GC in the standard
925 generation copying collector will usually require 3L bytes of
926 memory, where L is the amount of live data. This is because by
927 default (see the <option>+RTS -F</option> option) we allow the old
928 generation to grow to twice its size (2L) before collecting it, and
929 we require additionally L bytes to copy the live data into. When
930 using compacting collection (see the <option>+RTS -c</option>
931 option), this is reduced to 2L, and can further be reduced by
932 tweaking the <option>-F</option> option. Also add the size of the
933 allocation area (currently a fixed 512Kb).</para>
937 <para>The stack isn't counted in the heap profile by default. See the
938 <option>+RTS -xt</option> option.</para>
942 <para>The program text itself, the C stack, any non-heap data (eg. data
943 allocated by foreign libraries, and data allocated by the RTS), and
944 <literal>mmap()</literal>'d memory are not counted in the heap profile.</para>
952 <title><command>hp2ps</command>––heap profile to PostScript</title>
954 <indexterm><primary><command>hp2ps</command></primary></indexterm>
955 <indexterm><primary>heap profiles</primary></indexterm>
956 <indexterm><primary>postscript, from heap profiles</primary></indexterm>
957 <indexterm><primary><option>-h<break-down></option></primary></indexterm>
962 hp2ps [flags] [<file>[.hp]]
966 <command>hp2ps</command><indexterm><primary>hp2ps
967 program</primary></indexterm> converts a heap profile as produced
968 by the <option>-h<break-down></option> runtime option into a
969 PostScript graph of the heap profile. By convention, the file to
970 be processed by <command>hp2ps</command> has a
971 <filename>.hp</filename> extension. The PostScript output is
972 written to <filename><file>@.ps</filename>. If
973 <filename><file></filename> is omitted entirely, then the
974 program behaves as a filter.</para>
976 <para><command>hp2ps</command> is distributed in
977 <filename>ghc/utils/hp2ps</filename> in a GHC source
978 distribution. It was originally developed by Dave Wakeling as part
979 of the HBC/LML heap profiler.</para>
981 <para>The flags are:</para>
986 <term><option>-d</option></term>
988 <para>In order to make graphs more readable,
989 <command>hp2ps</command> sorts the shaded bands for each
990 identifier. The default sort ordering is for the bands with
991 the largest area to be stacked on top of the smaller ones.
992 The <option>-d</option> option causes rougher bands (those
993 representing series of values with the largest standard
994 deviations) to be stacked on top of smoother ones.</para>
999 <term><option>-b</option></term>
1001 <para>Normally, <command>hp2ps</command> puts the title of
1002 the graph in a small box at the top of the page. However, if
1003 the JOB string is too long to fit in a small box (more than
1004 35 characters), then <command>hp2ps</command> will choose to
1005 use a big box instead. The <option>-b</option> option
1006 forces <command>hp2ps</command> to use a big box.</para>
1011 <term><option>-e<float>[in|mm|pt]</option></term>
1013 <para>Generate encapsulated PostScript suitable for
1014 inclusion in LaTeX documents. Usually, the PostScript graph
1015 is drawn in landscape mode in an area 9 inches wide by 6
1016 inches high, and <command>hp2ps</command> arranges for this
1017 area to be approximately centred on a sheet of a4 paper.
1018 This format is convenient of studying the graph in detail,
1019 but it is unsuitable for inclusion in LaTeX documents. The
1020 <option>-e</option> option causes the graph to be drawn in
1021 portrait mode, with float specifying the width in inches,
1022 millimetres or points (the default). The resulting
1023 PostScript file conforms to the Encapsulated PostScript
1024 (EPS) convention, and it can be included in a LaTeX document
1025 using Rokicki's dvi-to-PostScript converter
1026 <command>dvips</command>.</para>
1031 <term><option>-g</option></term>
1033 <para>Create output suitable for the <command>gs</command>
1034 PostScript previewer (or similar). In this case the graph is
1035 printed in portrait mode without scaling. The output is
1036 unsuitable for a laser printer.</para>
1041 <term><option>-l</option></term>
1043 <para>Normally a profile is limited to 20 bands with
1044 additional identifiers being grouped into an
1045 <literal>OTHER</literal> band. The <option>-l</option> flag
1046 removes this 20 band and limit, producing as many bands as
1047 necessary. No key is produced as it won't fit!. It is useful
1048 for creation time profiles with many bands.</para>
1053 <term><option>-m<int></option></term>
1055 <para>Normally a profile is limited to 20 bands with
1056 additional identifiers being grouped into an
1057 <literal>OTHER</literal> band. The <option>-m</option> flag
1058 specifies an alternative band limit (the maximum is
1061 <para><option>-m0</option> requests the band limit to be
1062 removed. As many bands as necessary are produced. However no
1063 key is produced as it won't fit! It is useful for displaying
1064 creation time profiles with many bands.</para>
1069 <term><option>-p</option></term>
1071 <para>Use previous parameters. By default, the PostScript
1072 graph is automatically scaled both horizontally and
1073 vertically so that it fills the page. However, when
1074 preparing a series of graphs for use in a presentation, it
1075 is often useful to draw a new graph using the same scale,
1076 shading and ordering as a previous one. The
1077 <option>-p</option> flag causes the graph to be drawn using
1078 the parameters determined by a previous run of
1079 <command>hp2ps</command> on <filename>file</filename>. These
1080 are extracted from <filename>file@.aux</filename>.</para>
1085 <term><option>-s</option></term>
1087 <para>Use a small box for the title.</para>
1092 <term><option>-t<float></option></term>
1094 <para>Normally trace elements which sum to a total of less
1095 than 1% of the profile are removed from the
1096 profile. The <option>-t</option> option allows this
1097 percentage to be modified (maximum 5%).</para>
1099 <para><option>-t0</option> requests no trace elements to be
1100 removed from the profile, ensuring that all the data will be
1106 <term><option>-c</option></term>
1108 <para>Generate colour output.</para>
1113 <term><option>-y</option></term>
1115 <para>Ignore marks.</para>
1120 <term><option>-?</option></term>
1122 <para>Print out usage information.</para>
1128 <sect2 id="manipulating-hp">
1129 <title>Manipulating the hp file</title>
1131 <para>(Notes kindly offered by Jan-Willhem Maessen.)</para>
1134 The <filename>FOO.hp</filename> file produced when you ask for the
1135 heap profile of a program <filename>FOO</filename> is a text file with a particularly
1136 simple structure. Here's a representative example, with much of the
1137 actual data omitted:
1140 DATE "Thu Dec 26 18:17 2002"
1141 SAMPLE_UNIT "seconds"
1152 BEGIN_SAMPLE 11695.47
1155 The first four lines (<literal>JOB</literal>, <literal>DATE</literal>, <literal>SAMPLE_UNIT</literal>, <literal>VALUE_UNIT</literal>) form a
1156 header. Each block of lines starting with <literal>BEGIN_SAMPLE</literal> and ending
1157 with <literal>END_SAMPLE</literal> forms a single sample (you can think of this as a
1158 vertical slice of your heap profile). The hp2ps utility should accept
1159 any input with a properly-formatted header followed by a series of
1165 <title>Zooming in on regions of your profile</title>
1168 You can look at particular regions of your profile simply by loading a
1169 copy of the <filename>.hp</filename> file into a text editor and deleting the unwanted
1170 samples. The resulting <filename>.hp</filename> file can be run through <command>hp2ps</command> and viewed
1176 <title>Viewing the heap profile of a running program</title>
1179 The <filename>.hp</filename> file is generated incrementally as your
1180 program runs. In principle, running <command>hp2ps</command> on the incomplete file
1181 should produce a snapshot of your program's heap usage. However, the
1182 last sample in the file may be incomplete, causing <command>hp2ps</command> to fail. If
1183 you are using a machine with UNIX utilities installed, it's not too
1184 hard to work around this problem (though the resulting command line
1185 looks rather Byzantine):
1187 head -`fgrep -n END_SAMPLE FOO.hp | tail -1 | cut -d : -f 1` FOO.hp \
1191 The command <command>fgrep -n END_SAMPLE FOO.hp</command> finds the
1192 end of every complete sample in <filename>FOO.hp</filename>, and labels each sample with
1193 its ending line number. We then select the line number of the last
1194 complete sample using <command>tail</command> and <command>cut</command>. This is used as a
1195 parameter to <command>head</command>; the result is as if we deleted the final
1196 incomplete sample from <filename>FOO.hp</filename>. This results in a properly-formatted
1197 .hp file which we feed directly to <command>hp2ps</command>.
1201 <title>Viewing a heap profile in real time</title>
1204 The <command>gv</command> and <command>ghostview</command> programs
1205 have a "watch file" option can be used to view an up-to-date heap
1206 profile of your program as it runs. Simply generate an incremental
1207 heap profile as described in the previous section. Run <command>gv</command> on your
1210 gv -watch -seascape FOO.ps
1212 If you forget the <literal>-watch</literal> flag you can still select
1213 "Watch file" from the "State" menu. Now each time you generate a new
1214 profile <filename>FOO.ps</filename> the view will update automatically.
1218 This can all be encapsulated in a little script:
1221 head -`fgrep -n END_SAMPLE FOO.hp | tail -1 | cut -d : -f 1` FOO.hp \
1223 gv -watch -seascape FOO.ps &
1225 sleep 10 # We generate a new profile every 10 seconds.
1226 head -`fgrep -n END_SAMPLE FOO.hp | tail -1 | cut -d : -f 1` FOO.hp \
1230 Occasionally <command>gv</command> will choke as it tries to read an incomplete copy of
1231 <filename>FOO.ps</filename> (because <command>hp2ps</command> is still running as an update
1232 occurs). A slightly more complicated script works around this
1233 problem, by using the fact that sending a SIGHUP to gv will cause it
1234 to re-read its input file:
1237 head -`fgrep -n END_SAMPLE FOO.hp | tail -1 | cut -d : -f 1` FOO.hp \
1243 head -`fgrep -n END_SAMPLE FOO.hp | tail -1 | cut -d : -f 1` FOO.hp \
1253 <title>Observing Code Coverage</title>
1254 <indexterm><primary>code coverage</primary></indexterm>
1255 <indexterm><primary>Haskell Program Coverage</primary></indexterm>
1256 <indexterm><primary>hpc</primary></indexterm>
1259 Code coverage tools allow a programmer to determine what parts of
1260 their code have been actually executed, and which parts have
1261 never actually been invoked. GHC has an option for generating
1262 instrumented code that records code coverage as part of the
1263 <ulink url="http://www.haskell.org/hpc">Haskell Program Coverage
1264 </ulink>(HPC) toolkit, which is included with GHC. HPC tools can
1265 be used to render the generated code coverage information into
1266 human understandable format. </para>
1269 Correctly instrumented code provides coverage information of two
1270 kinds: source coverage and boolean-control coverage. Source
1271 coverage is the extent to which every part of the program was
1272 used, measured at three different levels: declarations (both
1273 top-level and local), alternatives (among several equations or
1274 case branches) and expressions (at every level). Boolean
1275 coverage is the extent to which each of the values True and
1276 False is obtained in every syntactic boolean context (ie. guard,
1277 condition, qualifier). </para>
1280 HPC displays both kinds of information in two primary ways:
1281 textual reports with summary statistics (hpc report) and sources
1282 with color mark-up (hpc markup). For boolean coverage, there
1283 are four possible outcomes for each guard, condition or
1284 qualifier: both True and False values occur; only True; only
1285 False; never evaluated. In hpc-markup output, highlighting with
1286 a yellow background indicates a part of the program that was
1287 never evaluated; a green background indicates an always-True
1288 expression and a red background indicates an always-False one.
1291 <sect2><title>A small example: Reciprocation</title>
1294 For an example we have a program, called Recip.hs, which computes exact decimal
1295 representations of reciprocals, with recurring parts indicated in
1299 reciprocal :: Int -> (String, Int)
1300 reciprocal n | n > 1 = ('0' : '.' : digits, recur)
1302 "attempting to compute reciprocal of number <= 1"
1304 (digits, recur) = divide n 1 []
1305 divide :: Int -> Int -> [Int] -> (String, Int)
1306 divide n c cs | c `elem` cs = ([], position c cs)
1307 | r == 0 = (show q, 0)
1308 | r /= 0 = (show q ++ digits, recur)
1310 (q, r) = (c*10) `quotRem` n
1311 (digits, recur) = divide n r (c:cs)
1313 position :: Int -> [Int] -> Int
1314 position n (x:xs) | n==x = 1
1315 | otherwise = 1 + position n xs
1317 showRecip :: Int -> String
1319 "1/" ++ show n ++ " = " ++
1320 if r==0 then d else take p d ++ "(" ++ drop p d ++ ")"
1323 (d, r) = reciprocal n
1327 putStrLn (showRecip number)
1331 <para>The HPC instrumentation is enabled using the -fhpc flag.
1335 $ ghc -fhpc Recip.hs --make
1337 <para>HPC index (.mix) files are placed placed in .hpc subdirectory. These can be considered like
1338 the .hi files for HPC.
1345 <para>We can generate a textual summary of coverage:</para>
1348 80% expressions used (81/101)
1349 12% boolean coverage (1/8)
1350 14% guards (1/7), 3 always True,
1353 0% 'if' conditions (0/1), 1 always False
1354 100% qualifiers (0/0)
1355 55% alternatives used (5/9)
1356 100% local declarations used (9/9)
1357 100% top-level declarations used (5/5)
1359 <para>We can also generate a marked-up version of the source.</para>
1362 writing Recip.hs.html
1365 This generates one file per Haskell module, and 4 index files,
1366 hpc_index.html, hpc_index_alt.html, hpc_index_exp.html,
1371 <sect2><title>Options for instrumenting code for coverage</title>
1373 Turning on code coverage is easy, use the -fhpc flag.
1374 Instrumented and non-instrumented can be freely mixed.
1375 When compiling the Main module GHC automatically detects when there
1376 is an hpc compiled file, and adds the correct initialization code.
1381 <sect2><title>The hpc toolkit</title>
1384 The hpc toolkit uses a cvs/svn/darcs-like interface, where a
1385 single binary contains many function units.</para>
1388 Usage: hpc COMMAND ...
1391 help Display help for hpc or a single command
1393 report Output textual report about program coverage
1394 markup Markup Haskell source with program coverage
1395 Processing Coverage files:
1396 sum Sum multiple .tix files in a single .tix file
1397 combine Combine two .tix files in a single .tix file
1398 map Map a function over a single .tix file
1400 overlay Generate a .tix file from an overlay file
1401 draft Generate draft overlay that provides 100% coverage
1403 show Show .tix file in readable, verbose format
1404 version Display version for hpc
1407 <para>In general, these options act on .tix file after an
1408 instrumented binary has generated it, which hpc acting as a
1409 conduit between the raw .tix file, and the more detailed reports
1414 The hpc tool assumes you are in the top-level directory of
1415 the location where you built your application, and the .tix
1416 file is in the same top-level directory. You can use the
1417 flag --srcdir to use hpc for any other directory, and use
1418 --srcdir multiple times to analyse programs compiled from
1419 difference locations, as is typical for packages.
1423 We now explain in more details the major modes of hpc.
1426 <sect3><title>hpc report</title>
1427 <para>hpc report gives a textual report of coverage. By default,
1428 all modules and packages are considered in generating report,
1429 unless include or exclude are used. The report is a summary
1430 unless the --per-module flag is used. The --xml-output option
1431 allows for tools to use hpc to glean coverage.
1435 Usage: hpc report [OPTION] .. <TIX_FILE> [<MODULE> [<MODULE> ..]]
1439 --per-module show module level detail
1440 --decl-list show unused decls
1441 --exclude=[PACKAGE:][MODULE] exclude MODULE and/or PACKAGE
1442 --include=[PACKAGE:][MODULE] include MODULE and/or PACKAGE
1443 --srcdir=DIR path to source directory of .hs files
1444 multi-use of srcdir possible
1445 --hpcdir=DIR sub-directory that contains .mix files
1446 default .hpc [rarely used]
1447 --xml-output show output in XML
1450 <sect3><title>hpc markup</title>
1451 <para>hpc markup marks up source files into colored html.
1455 Usage: hpc markup [OPTION] .. <TIX_FILE> [<MODULE> [<MODULE> ..]]
1459 --exclude=[PACKAGE:][MODULE] exclude MODULE and/or PACKAGE
1460 --include=[PACKAGE:][MODULE] include MODULE and/or PACKAGE
1461 --srcdir=DIR path to source directory of .hs files
1462 multi-use of srcdir possible
1463 --hpcdir=DIR sub-directory that contains .mix files
1464 default .hpc [rarely used]
1465 --fun-entry-count show top-level function entry counts
1466 --highlight-covered highlight covered code, rather that code gaps
1467 --destdir=DIR path to write output to
1471 <sect3><title>hpc sum</title>
1472 <para>hpc sum adds together any number of .tix files into a single
1473 .tix file. hpc sum does not change the original .tix file; it generates a new .tix file.
1477 Usage: hpc sum [OPTION] .. <TIX_FILE> [<TIX_FILE> [<TIX_FILE> ..]]
1478 Sum multiple .tix files in a single .tix file
1482 --exclude=[PACKAGE:][MODULE] exclude MODULE and/or PACKAGE
1483 --include=[PACKAGE:][MODULE] include MODULE and/or PACKAGE
1484 --output=FILE output FILE
1485 --union use the union of the module namespace (default is intersection)
1488 <sect3><title>hpc combine</title>
1489 <para>hpc combine is the swiss army knife of hpc. It can be
1490 used to take the difference between .tix files, to subtract one
1491 .tix file from another, or to add two .tix files. hpc combine does not
1492 change the original .tix file; it generates a new .tix file.
1496 Usage: hpc combine [OPTION] .. <TIX_FILE> <TIX_FILE>
1497 Combine two .tix files in a single .tix file
1501 --exclude=[PACKAGE:][MODULE] exclude MODULE and/or PACKAGE
1502 --include=[PACKAGE:][MODULE] include MODULE and/or PACKAGE
1503 --output=FILE output FILE
1504 --function=FUNCTION combine .tix files with join function, default = ADD
1505 FUNCTION = ADD | DIFF | SUB
1506 --union use the union of the module namespace (default is intersection)
1509 <sect3><title>hpc map</title>
1510 <para>hpc map inverts or zeros a .tix file. hpc map does not
1511 change the original .tix file; it generates a new .tix file.
1515 Usage: hpc map [OPTION] .. <TIX_FILE>
1516 Map a function over a single .tix file
1520 --exclude=[PACKAGE:][MODULE] exclude MODULE and/or PACKAGE
1521 --include=[PACKAGE:][MODULE] include MODULE and/or PACKAGE
1522 --output=FILE output FILE
1523 --function=FUNCTION apply function to .tix files, default = ID
1524 FUNCTION = ID | INV | ZERO
1525 --union use the union of the module namespace (default is intersection)
1528 <sect3><title>hpc overlay and hpc draft</title>
1530 Overlays are an experimental feature of HPC, a textual description
1531 of coverage. hpc draft is used to generate a draft overlay from a .tix file,
1532 and hpc overlay generates a .tix files from an overlay.
1536 Usage: hpc overlay [OPTION] .. <OVERLAY_FILE> [<OVERLAY_FILE> [...]]
1540 --srcdir=DIR path to source directory of .hs files
1541 multi-use of srcdir possible
1542 --hpcdir=DIR sub-directory that contains .mix files
1543 default .hpc [rarely used]
1544 --output=FILE output FILE
1546 Usage: hpc draft [OPTION] .. <TIX_FILE>
1550 --exclude=[PACKAGE:][MODULE] exclude MODULE and/or PACKAGE
1551 --include=[PACKAGE:][MODULE] include MODULE and/or PACKAGE
1552 --srcdir=DIR path to source directory of .hs files
1553 multi-use of srcdir possible
1554 --hpcdir=DIR sub-directory that contains .mix files
1555 default .hpc [rarely used]
1556 --output=FILE output FILE
1560 <sect2><title>Caveats and Shortcomings of Haskell Program Coverage</title>
1562 HPC does not attempt to lock the .tix file, so multiple concurrently running
1563 binaries in the same directory will exhibit a race condition. There is no way
1564 to change the name of the .tix file generated, apart from renaming the binary.
1565 HPC does not work with GHCi.
1570 <sect1 id="ticky-ticky">
1571 <title>Using “ticky-ticky” profiling (for implementors)</title>
1572 <indexterm><primary>ticky-ticky profiling</primary></indexterm>
1574 <para>(ToDo: document properly.)</para>
1576 <para>It is possible to compile Glasgow Haskell programs so that
1577 they will count lots and lots of interesting things, e.g., number
1578 of updates, number of data constructors entered, etc., etc. We
1579 call this “ticky-ticky”
1580 profiling,<indexterm><primary>ticky-ticky
1581 profiling</primary></indexterm> <indexterm><primary>profiling,
1582 ticky-ticky</primary></indexterm> because that's the sound a Sun4
1583 makes when it is running up all those counters
1584 (<emphasis>slowly</emphasis>).</para>
1586 <para>Ticky-ticky profiling is mainly intended for implementors;
1587 it is quite separate from the main “cost-centre”
1588 profiling system, intended for all users everywhere.</para>
1590 <para>To be able to use ticky-ticky profiling, you will need to
1591 have built the ticky RTS. (This should be described in
1592 the building guide, but amounts to building the RTS with way
1593 "t" enabled.)</para>
1595 <para>To get your compiled program to spit out the ticky-ticky
1596 numbers, use a <option>-r</option> RTS
1597 option<indexterm><primary>-r RTS option</primary></indexterm>.
1598 See <xref linkend="runtime-control"/>.</para>
1600 <para>Compiling your program with the <option>-ticky</option>
1601 switch yields an executable that performs these counts. Here is a
1602 sample ticky-ticky statistics file, generated by the invocation
1603 <command>foo +RTS -rfoo.ticky</command>.</para>
1606 foo +RTS -rfoo.ticky
1609 ALLOCATIONS: 3964631 (11330900 words total: 3999476 admin, 6098829 goods, 1232595 slop)
1610 total words: 2 3 4 5 6+
1611 69647 ( 1.8%) function values 50.0 50.0 0.0 0.0 0.0
1612 2382937 ( 60.1%) thunks 0.0 83.9 16.1 0.0 0.0
1613 1477218 ( 37.3%) data values 66.8 33.2 0.0 0.0 0.0
1614 0 ( 0.0%) big tuples
1615 2 ( 0.0%) black holes 0.0 100.0 0.0 0.0 0.0
1616 0 ( 0.0%) prim things
1617 34825 ( 0.9%) partial applications 0.0 0.0 0.0 100.0 0.0
1618 2 ( 0.0%) thread state objects 0.0 0.0 0.0 0.0 100.0
1620 Total storage-manager allocations: 3647137 (11882004 words)
1621 [551104 words lost to speculative heap-checks]
1625 ENTERS: 9400092 of which 2005772 (21.3%) direct to the entry code
1626 [the rest indirected via Node's info ptr]
1627 1860318 ( 19.8%) thunks
1628 3733184 ( 39.7%) data values
1629 3149544 ( 33.5%) function values
1630 [of which 1999880 (63.5%) bypassed arg-satisfaction chk]
1631 348140 ( 3.7%) partial applications
1632 308906 ( 3.3%) normal indirections
1633 0 ( 0.0%) permanent indirections
1636 2137257 ( 36.4%) from entering a new constructor
1637 [the rest from entering an existing constructor]
1638 2349219 ( 40.0%) vectored [the rest unvectored]
1640 RET_NEW: 2137257: 32.5% 46.2% 21.3% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0%
1641 RET_OLD: 3733184: 2.8% 67.9% 29.3% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0%
1642 RET_UNBOXED_TUP: 2: 0.0% 0.0%100.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0%
1644 RET_VEC_RETURN : 2349219: 0.0% 0.0%100.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0%
1646 UPDATE FRAMES: 2241725 (0 omitted from thunks)
1650 0 ( 0.0%) data values
1651 34827 ( 1.6%) partial applications
1652 [2 in place, 34825 allocated new space]
1653 2206898 ( 98.4%) updates to existing heap objects (46 by squeezing)
1654 UPD_CON_IN_NEW: 0: 0 0 0 0 0 0 0 0 0
1655 UPD_PAP_IN_NEW: 34825: 0 0 0 34825 0 0 0 0 0
1657 NEW GEN UPDATES: 2274700 ( 99.9%)
1659 OLD GEN UPDATES: 1852 ( 0.1%)
1661 Total bytes copied during GC: 190096
1663 **************************************************
1664 3647137 ALLOC_HEAP_ctr
1665 11882004 ALLOC_HEAP_tot
1670 34831 ALLOC_FUN_hst_0
1671 34816 ALLOC_FUN_hst_1
1675 2382937 ALLOC_UP_THK_ctr
1678 0 E!NT_PERM_IND_ctr requires +RTS -Z
1679 [... lots more info omitted ...]
1680 0 GC_SEL_ABANDONED_ctr
1683 0 GC_FAILED_PROMOTION_ctr
1684 47524 GC_WORDS_COPIED_ctr
1687 <para>The formatting of the information above the row of asterisks
1688 is subject to change, but hopefully provides a useful
1689 human-readable summary. Below the asterisks <emphasis>all
1690 counters</emphasis> maintained by the ticky-ticky system are
1691 dumped, in a format intended to be machine-readable: zero or more
1692 spaces, an integer, a space, the counter name, and a newline.</para>
1694 <para>In fact, not <emphasis>all</emphasis> counters are
1695 necessarily dumped; compile- or run-time flags can render certain
1696 counters invalid. In this case, either the counter will simply
1697 not appear, or it will appear with a modified counter name,
1698 possibly along with an explanation for the omission (notice
1699 <literal>ENT_PERM_IND_ctr</literal> appears
1700 with an inserted <literal>!</literal> above). Software analysing
1701 this output should always check that it has the counters it
1702 expects. Also, beware: some of the counters can have
1703 <emphasis>large</emphasis> values!</para>
1710 ;;; Local Variables: ***
1712 ;;; sgml-parent-document: ("users_guide.xml" "book" "chapter") ***