Gitiles
Code Review Sign In
gerrit-public.fairphone.software / platform / external / valgrind / a0664b9ca67b594bd6f570a61d3301167a24750c / . / docs / html / ms-manual.html
blob: fbb0e5a3961a2b3251609a002876d379e71370b8 [file] [log] [blame]
Elliott Hughesa0664b92017-04-18 17:46:52 -0700[diff] [blame^]1<html>
2<head>
3<meta http-equiv="Content-Type" content="text/html; charset=ISO-8859-1">
4<title>9. Massif: a heap profiler</title>
5<link rel="stylesheet" type="text/css" href="vg_basic.css">
6<meta name="generator" content="DocBook XSL Stylesheets V1.78.1">
7<link rel="home" href="index.html" title="Valgrind Documentation">
8<link rel="up" href="manual.html" title="Valgrind User Manual">
9<link rel="prev" href="drd-manual.html" title="8. DRD: a thread error detector">
10<link rel="next" href="dh-manual.html" title="10. DHAT: a dynamic heap analysis tool">
11</head>
12<body bgcolor="white" text="black" link="#0000FF" vlink="#840084" alink="#0000FF">
13<div><table class="nav" width="100%" cellspacing="3" cellpadding="3" border="0" summary="Navigation header"><tr>
14<td width="22px" align="center" valign="middle"><a accesskey="p" href="drd-manual.html"><img src="images/prev.png" width="18" height="21" border="0" alt="Prev"></a></td>
15<td width="25px" align="center" valign="middle"><a accesskey="u" href="manual.html"><img src="images/up.png" width="21" height="18" border="0" alt="Up"></a></td>
16<td width="31px" align="center" valign="middle"><a accesskey="h" href="index.html"><img src="images/home.png" width="27" height="20" border="0" alt="Up"></a></td>
17<th align="center" valign="middle">Valgrind User Manual</th>
18<td width="22px" align="center" valign="middle"><a accesskey="n" href="dh-manual.html"><img src="images/next.png" width="18" height="21" border="0" alt="Next"></a></td>
19</tr></table></div>
20<div class="chapter">
21<div class="titlepage"><div><div><h1 class="title">
22<a name="ms-manual"></a>9. Massif: a heap profiler</h1></div></div></div>
23<div class="toc">
24<p><b>Table of Contents</b></p>
25<dl class="toc">
26<dt><span class="sect1"><a href="ms-manual.html#ms-manual.overview">9.1. Overview</a></span></dt>
27<dt><span class="sect1"><a href="ms-manual.html#ms-manual.using">9.2. Using Massif and ms_print</a></span></dt>
28<dd><dl>
29<dt><span class="sect2"><a href="ms-manual.html#ms-manual.anexample">9.2.1. An Example Program</a></span></dt>
30<dt><span class="sect2"><a href="ms-manual.html#ms-manual.running-massif">9.2.2. Running Massif</a></span></dt>
31<dt><span class="sect2"><a href="ms-manual.html#ms-manual.running-ms_print">9.2.3. Running ms_print</a></span></dt>
32<dt><span class="sect2"><a href="ms-manual.html#ms-manual.theoutputpreamble">9.2.4. The Output Preamble</a></span></dt>
33<dt><span class="sect2"><a href="ms-manual.html#ms-manual.theoutputgraph">9.2.5. The Output Graph</a></span></dt>
34<dt><span class="sect2"><a href="ms-manual.html#ms-manual.thesnapshotdetails">9.2.6. The Snapshot Details</a></span></dt>
35<dt><span class="sect2"><a href="ms-manual.html#ms-manual.forkingprograms">9.2.7. Forking Programs</a></span></dt>
36<dt><span class="sect2"><a href="ms-manual.html#ms-manual.not-measured">9.2.8. Measuring All Memory in a Process</a></span></dt>
37<dt><span class="sect2"><a href="ms-manual.html#ms-manual.acting">9.2.9. Acting on Massif's Information</a></span></dt>
38</dl></dd>
39<dt><span class="sect1"><a href="ms-manual.html#ms-manual.options">9.3. Massif Command-line Options</a></span></dt>
40<dt><span class="sect1"><a href="ms-manual.html#ms-manual.monitor-commands">9.4. Massif Monitor Commands</a></span></dt>
41<dt><span class="sect1"><a href="ms-manual.html#ms-manual.clientreqs">9.5. Massif Client Requests</a></span></dt>
42<dt><span class="sect1"><a href="ms-manual.html#ms-manual.ms_print-options">9.6. ms_print Command-line Options</a></span></dt>
43<dt><span class="sect1"><a href="ms-manual.html#ms-manual.fileformat">9.7. Massif's Output File Format</a></span></dt>
44</dl>
45</div>
46<p>To use this tool, you must specify
47<code class="option">--tool=massif</code> on the Valgrind
48command line.</p>
49<div class="sect1">
50<div class="titlepage"><div><div><h2 class="title" style="clear: both">
51<a name="ms-manual.overview"></a>9.1. Overview</h2></div></div></div>
52<p>Massif is a heap profiler. It measures how much heap memory your
53program uses. This includes both the useful space, and the extra bytes
54allocated for book-keeping and alignment purposes. It can also
55measure the size of your program's stack(s), although it does not do so by
56default.</p>
57<p>Heap profiling can help you reduce the amount of memory your program
58uses. On modern machines with virtual memory, this provides the following
59benefits:</p>
60<div class="itemizedlist"><ul class="itemizedlist" style="list-style-type: disc; ">
61<li class="listitem"><p>It can speed up your program -- a smaller
62 program will interact better with your machine's caches and
63 avoid paging.</p></li>
64<li class="listitem"><p>If your program uses lots of memory, it will
65 reduce the chance that it exhausts your machine's swap
66 space.</p></li>
67</ul></div>
68<p>Also, there are certain space leaks that aren't detected by
69traditional leak-checkers, such as Memcheck's. That's because
70the memory isn't ever actually lost -- a pointer remains to it --
71but it's not in use. Programs that have leaks like this can
72unnecessarily increase the amount of memory they are using over
73time. Massif can help identify these leaks.</p>
74<p>Importantly, Massif tells you not only how much heap memory your
75program is using, it also gives very detailed information that indicates
76which parts of your program are responsible for allocating the heap memory.
77</p>
78</div>
79<div class="sect1">
80<div class="titlepage"><div><div><h2 class="title" style="clear: both">
81<a name="ms-manual.using"></a>9.2. Using Massif and ms_print</h2></div></div></div>
82<p>First off, as for the other Valgrind tools, you should compile with
83debugging info (the <code class="option">-g</code> option). It shouldn't
84matter much what optimisation level you compile your program with, as this
85is unlikely to affect the heap memory usage.</p>
86<p>Then, you need to run Massif itself to gather the profiling
87information, and then run ms_print to present it in a readable way.</p>
88<div class="sect2">
89<div class="titlepage"><div><div><h3 class="title">
90<a name="ms-manual.anexample"></a>9.2.1. An Example Program</h3></div></div></div>
91<p>An example will make things clear. Consider the following C program
92(annotated with line numbers) which allocates a number of different blocks
93on the heap.</p>
94<pre class="screen">
95 1 #include &lt;stdlib.h&gt;
96 2
97 3 void g(void)
98 4 {
99 5 malloc(4000);
100 6 }
101 7
102 8 void f(void)
103 9 {
10410 malloc(2000);
10511 g();
10612 }
10713
10814 int main(void)
10915 {
11016 int i;
11117 int* a[10];
11218
11319 for (i = 0; i &lt; 10; i++) {
11420 a[i] = malloc(1000);
11521 }
11622
11723 f();
11824
11925 g();
12026
12127 for (i = 0; i &lt; 10; i++) {
12228 free(a[i]);
12329 }
12430
12531 return 0;
12632 }
127</pre>
128</div>
129<div class="sect2">
130<div class="titlepage"><div><div><h3 class="title">
131<a name="ms-manual.running-massif"></a>9.2.2. Running Massif</h3></div></div></div>
132<p>To gather heap profiling information about the program
133<code class="computeroutput">prog</code>, type:</p>
134<pre class="screen">
135valgrind --tool=massif prog
136</pre>
137<p>The program will execute (slowly). Upon completion, no summary
138statistics are printed to Valgrind's commentary; all of Massif's profiling
139data is written to a file. By default, this file is called
140<code class="filename">massif.out.&lt;pid&gt;</code>, where
141<code class="filename">&lt;pid&gt;</code> is the process ID, although this filename
142can be changed with the <code class="option">--massif-out-file</code> option.</p>
143</div>
144<div class="sect2">
145<div class="titlepage"><div><div><h3 class="title">
146<a name="ms-manual.running-ms_print"></a>9.2.3. Running ms_print</h3></div></div></div>
147<p>To see the information gathered by Massif in an easy-to-read form, use
148ms_print. If the output file's name is
149<code class="filename">massif.out.12345</code>, type:</p>
150<pre class="screen">
151ms_print massif.out.12345</pre>
152<p>ms_print will produce (a) a graph showing the memory consumption over
153the program's execution, and (b) detailed information about the responsible
154allocation sites at various points in the program, including the point of
155peak memory allocation. The use of a separate script for presenting the
156results is deliberate: it separates the data gathering from its
157presentation, and means that new methods of presenting the data can be added in
158the future.</p>
159</div>
160<div class="sect2">
161<div class="titlepage"><div><div><h3 class="title">
162<a name="ms-manual.theoutputpreamble"></a>9.2.4. The Output Preamble</h3></div></div></div>
163<p>After running this program under Massif, the first part of ms_print's
164output contains a preamble which just states how the program, Massif and
165ms_print were each invoked:</p>
166<pre class="screen">
167--------------------------------------------------------------------------------
168Command: example
169Massif arguments: (none)
170ms_print arguments: massif.out.12797
171--------------------------------------------------------------------------------
172</pre>
173</div>
174<div class="sect2">
175<div class="titlepage"><div><div><h3 class="title">
176<a name="ms-manual.theoutputgraph"></a>9.2.5. The Output Graph</h3></div></div></div>
177<p>The next part is the graph that shows how memory consumption occurred
178as the program executed:</p>
179<pre class="screen">
180 KB
18119.63^ #
182 | #
183 | #
184 | #
185 | #
186 | #
187 | #
188 | #
189 | #
190 | #
191 | #
192 | #
193 | #
194 | #
195 | #
196 | #
197 | #
198 | :#
199 | :#
200 | :#
201 0 +-----------------------------------------------------------------------&gt;ki 0 113.4
202
203
204Number of snapshots: 25
205 Detailed snapshots: [9, 14 (peak), 24]
206</pre>
207<p>Why is most of the graph empty, with only a couple of bars at the very
208end? By default, Massif uses "instructions executed" as the unit of time.
209For very short-run programs such as the example, most of the executed
210instructions involve the loading and dynamic linking of the program. The
211execution of <code class="computeroutput">main</code> (and thus the heap
212allocations) only occur at the very end. For a short-running program like
213this, we can use the <code class="option">--time-unit=B</code> option
214to specify that we want the time unit to instead be the number of bytes
215allocated/deallocated on the heap and stack(s).</p>
216<p>If we re-run the program under Massif with this option, and then
217re-run ms_print, we get this more useful graph:</p>
218<pre class="screen">
21919.63^ ###
220 | #
221 | # ::
222 | # : :::
223 | :::::::::# : : ::
224 | : # : : : ::
225 | : # : : : : :::
226 | : # : : : : : ::
227 | ::::::::::: # : : : : : : :::
228 | : : # : : : : : : : ::
229 | ::::: : # : : : : : : : : ::
230 | @@@: : : # : : : : : : : : : @
231 | ::@ : : : # : : : : : : : : : @
232 | :::: @ : : : # : : : : : : : : : @
233 | ::: : @ : : : # : : : : : : : : : @
234 | ::: : : @ : : : # : : : : : : : : : @
235 | :::: : : : @ : : : # : : : : : : : : : @
236 | ::: : : : : @ : : : # : : : : : : : : : @
237 | :::: : : : : : @ : : : # : : : : : : : : : @
238 | ::: : : : : : : @ : : : # : : : : : : : : : @
239 0 +-----------------------------------------------------------------------&gt;KB 0 29.48
240
241Number of snapshots: 25
242 Detailed snapshots: [9, 14 (peak), 24]
243</pre>
244<p>The size of the graph can be changed with ms_print's
245<code class="option">--x</code> and <code class="option">--y</code> options. Each vertical bar
246represents a snapshot, i.e. a measurement of the memory usage at a certain
247point in time. If the next snapshot is more than one column away, a
248horizontal line of characters is drawn from the top of the snapshot to just
249before the next snapshot column. The text at the bottom show that 25
250snapshots were taken for this program, which is one per heap
251allocation/deallocation, plus a couple of extras. Massif starts by taking
252snapshots for every heap allocation/deallocation, but as a program runs for
253longer, it takes snapshots less frequently. It also discards older
254snapshots as the program goes on; when it reaches the maximum number of
255snapshots (100 by default, although changeable with the
256<code class="option">--max-snapshots</code> option) half of them are
257deleted. This means that a reasonable number of snapshots are always
258maintained.</p>
259<p>Most snapshots are <span class="emphasis"><em>normal</em></span>, and only basic
260information is recorded for them. Normal snapshots are represented in the
261graph by bars consisting of ':' characters.</p>
262<p>Some snapshots are <span class="emphasis"><em>detailed</em></span>. Information about
263where allocations happened are recorded for these snapshots, as we will see
264shortly. Detailed snapshots are represented in the graph by bars consisting
265of '@' characters. The text at the bottom show that 3 detailed
266snapshots were taken for this program (snapshots 9, 14 and 24). By default,
267every 10th snapshot is detailed, although this can be changed via the
268<code class="option">--detailed-freq</code> option.</p>
269<p>Finally, there is at most one <span class="emphasis"><em>peak</em></span> snapshot. The
270peak snapshot is a detailed snapshot, and records the point where memory
271consumption was greatest. The peak snapshot is represented in the graph by
272a bar consisting of '#' characters. The text at the bottom shows
273that snapshot 14 was the peak.</p>
274<p>Massif's determination of when the peak occurred can be wrong, for
275two reasons.</p>
276<div class="itemizedlist"><ul class="itemizedlist" style="list-style-type: disc; ">
277<li class="listitem"><p>Peak snapshots are only ever taken after a deallocation
278 happens. This avoids lots of unnecessary peak snapshot recordings
279 (imagine what happens if your program allocates a lot of heap blocks in
280 succession, hitting a new peak every time). But it means that if your
281 program never deallocates any blocks, no peak will be recorded. It also
282 means that if your program does deallocate blocks but later allocates to a
283 higher peak without subsequently deallocating, the reported peak will be
284 too low.
285 </p></li>
286<li class="listitem"><p>Even with this behaviour, recording the peak accurately
287 is slow. So by default Massif records a peak whose size is within 1% of
288 the size of the true peak. This inaccuracy in the peak measurement can be
289 changed with the <code class="option">--peak-inaccuracy</code> option.</p></li>
290</ul></div>
291<p>The following graph is from an execution of Konqueror, the KDE web
292browser. It shows what graphs for larger programs look like.</p>
293<pre class="screen">
294 MB
2953.952^ #
296 | @#:
297 | :@@#:
298 | @@::::@@#:
299 | @ :: :@@#::
300 | @@@ :: :@@#::
301 | @@:@@@ :: :@@#::
302 | :::@ :@@@ :: :@@#::
303 | : :@ :@@@ :: :@@#::
304 | :@: :@ :@@@ :: :@@#::
305 | @@:@: :@ :@@@ :: :@@#:::
306 | : :: ::@@:@: :@ :@@@ :: :@@#:::
307 | :@@: ::::: ::::@@@:::@@:@: :@ :@@@ :: :@@#:::
308 | ::::@@: ::: ::::::: @ :::@@:@: :@ :@@@ :: :@@#:::
309 | @: ::@@: ::: ::::::: @ :::@@:@: :@ :@@@ :: :@@#:::
310 | @: ::@@: ::: ::::::: @ :::@@:@: :@ :@@@ :: :@@#:::
311 | @: ::@@:::::: ::::::: @ :::@@:@: :@ :@@@ :: :@@#:::
312 | ::@@@: ::@@:: ::: ::::::: @ :::@@:@: :@ :@@@ :: :@@#:::
313 | :::::@ @: ::@@:: ::: ::::::: @ :::@@:@: :@ :@@@ :: :@@#:::
314 | @@:::::@ @: ::@@:: ::: ::::::: @ :::@@:@: :@ :@@@ :: :@@#:::
315 0 +-----------------------------------------------------------------------&gt;Mi
316 0 626.4
317
318Number of snapshots: 63
319 Detailed snapshots: [3, 4, 10, 11, 15, 16, 29, 33, 34, 36, 39, 41,
320 42, 43, 44, 49, 50, 51, 53, 55, 56, 57 (peak)]
321</pre>
322<p>Note that the larger size units are KB, MB, GB, etc. As is typical
323for memory measurements, these are based on a multiplier of 1024, rather
324than the standard SI multiplier of 1000. Strictly speaking, they should be
325written KiB, MiB, GiB, etc.</p>
326</div>
327<div class="sect2">
328<div class="titlepage"><div><div><h3 class="title">
329<a name="ms-manual.thesnapshotdetails"></a>9.2.6. The Snapshot Details</h3></div></div></div>
330<p>Returning to our example, the graph is followed by the detailed
331information for each snapshot. The first nine snapshots are normal, so only
332a small amount of information is recorded for each one:</p>
333<pre class="screen">
334--------------------------------------------------------------------------------
335 n time(B) total(B) useful-heap(B) extra-heap(B) stacks(B)
336--------------------------------------------------------------------------------
337 0 0 0 0 0 0
338 1 1,008 1,008 1,000 8 0
339 2 2,016 2,016 2,000 16 0
340 3 3,024 3,024 3,000 24 0
341 4 4,032 4,032 4,000 32 0
342 5 5,040 5,040 5,000 40 0
343 6 6,048 6,048 6,000 48 0
344 7 7,056 7,056 7,000 56 0
345 8 8,064 8,064 8,000 64 0
346</pre>
347<p>Each normal snapshot records several things.</p>
348<div class="itemizedlist"><ul class="itemizedlist" style="list-style-type: disc; ">
349<li class="listitem"><p>Its number.</p></li>
350<li class="listitem"><p>The time it was taken. In this case, the time unit is
351 bytes, due to the use of
352 <code class="option">--time-unit=B</code>.</p></li>
353<li class="listitem"><p>The total memory consumption at that point.</p></li>
354<li class="listitem"><p>The number of useful heap bytes allocated at that point.
355 This reflects the number of bytes asked for by the
356 program.</p></li>
357<li class="listitem">
358<p>The number of extra heap bytes allocated at that point.
359 This reflects the number of bytes allocated in excess of what the program
360 asked for. There are two sources of extra heap bytes.</p>
361<p>First, every heap block has administrative bytes associated with it.
362 The exact number of administrative bytes depends on the details of the
363 allocator. By default Massif assumes 8 bytes per block, as can be seen
364 from the example, but this number can be changed via the
365 <code class="option">--heap-admin</code> option.</p>
366<p>Second, allocators often round up the number of bytes asked for to a
367 larger number, usually 8 or 16. This is required to ensure that elements
368 within the block are suitably aligned. If N bytes are asked for, Massif
369 rounds N up to the nearest multiple of the value specified by the
370 <code class="option"><a class="xref" href="manual-core.html#opt.alignment">--alignment</a></code> option.
371 </p>
372</li>
373<li class="listitem"><p>The size of the stack(s). By default, stack profiling is
374 off as it slows Massif down greatly. Therefore, the stack column is zero
375 in the example. Stack profiling can be turned on with the
376 <code class="option">--stacks=yes</code> option.
377
378 </p></li>
379</ul></div>
380<p>The next snapshot is detailed. As well as the basic counts, it gives
381an allocation tree which indicates exactly which pieces of code were
382responsible for allocating heap memory:</p>
383<pre class="screen">
384 9 9,072 9,072 9,000 72 0
38599.21% (9,000B) (heap allocation functions) malloc/new/new[], --alloc-fns, etc.
386-&gt;99.21% (9,000B) 0x804841A: main (example.c:20)
387</pre>
388<p>The allocation tree can be read from the top down. The first line
389indicates all heap allocation functions such as <code class="function">malloc</code>
390and C++ <code class="function">new</code>. All heap allocations go through these
391functions, and so all 9,000 useful bytes (which is 99.21% of all allocated
392bytes) go through them. But how were <code class="function">malloc</code> and new
393called? At this point, every allocation so far has been due to line 20
394inside <code class="function">main</code>, hence the second line in the tree. The
395<code class="option">-&gt;</code> indicates that main (line 20) called
396<code class="function">malloc</code>.</p>
397<p>Let's see what the subsequent output shows happened next:</p>
398<pre class="screen">
399--------------------------------------------------------------------------------
400 n time(B) total(B) useful-heap(B) extra-heap(B) stacks(B)
401--------------------------------------------------------------------------------
402 10 10,080 10,080 10,000 80 0
403 11 12,088 12,088 12,000 88 0
404 12 16,096 16,096 16,000 96 0
405 13 20,104 20,104 20,000 104 0
406 14 20,104 20,104 20,000 104 0
40799.48% (20,000B) (heap allocation functions) malloc/new/new[], --alloc-fns, etc.
408-&gt;49.74% (10,000B) 0x804841A: main (example.c:20)
409|
410-&gt;39.79% (8,000B) 0x80483C2: g (example.c:5)
411| -&gt;19.90% (4,000B) 0x80483E2: f (example.c:11)
412| | -&gt;19.90% (4,000B) 0x8048431: main (example.c:23)
413| |
414| -&gt;19.90% (4,000B) 0x8048436: main (example.c:25)
415|
416-&gt;09.95% (2,000B) 0x80483DA: f (example.c:10)
417 -&gt;09.95% (2,000B) 0x8048431: main (example.c:23)
418</pre>
419<p>The first four snapshots are similar to the previous ones. But then
420the global allocation peak is reached, and a detailed snapshot (number 14)
421is taken. Its allocation tree shows that 20,000B of useful heap memory has
422been allocated, and the lines and arrows indicate that this is from three
423different code locations: line 20, which is responsible for 10,000B
424(49.74%); line 5, which is responsible for 8,000B (39.79%); and line 10,
425which is responsible for 2,000B (9.95%).</p>
426<p>We can then drill down further in the allocation tree. For example,
427of the 8,000B asked for by line 5, half of it was due to a call from line
42811, and half was due to a call from line 25.</p>
429<p>In short, Massif collates the stack trace of every single allocation
430point in the program into a single tree, which gives a complete picture at
431a particular point in time of how and why all heap memory was
432allocated.</p>
433<p>Note that the tree entries correspond not to functions, but to
434individual code locations. For example, if function <code class="function">A</code>
435calls <code class="function">malloc</code>, and function <code class="function">B</code> calls
436<code class="function">A</code> twice, once on line 10 and once on line 11, then
437the two calls will result in two distinct stack traces in the tree. In
438contrast, if <code class="function">B</code> calls <code class="function">A</code> repeatedly
439from line 15 (e.g. due to a loop), then each of those calls will be
440represented by the same stack trace in the tree.</p>
441<p>Note also that each tree entry with children in the example satisfies an
442invariant: the entry's size is equal to the sum of its children's sizes.
443For example, the first entry has size 20,000B, and its children have sizes
44410,000B, 8,000B, and 2,000B. In general, this invariant almost always
445holds. However, in rare circumstances stack traces can be malformed, in
446which case a stack trace can be a sub-trace of another stack trace. This
447means that some entries in the tree may not satisfy the invariant -- the
448entry's size will be greater than the sum of its children's sizes. This is
449not a big problem, but could make the results confusing. Massif can
450sometimes detect when this happens; if it does, it issues a warning:</p>
451<pre class="screen">
452Warning: Malformed stack trace detected. In Massif's output,
453 the size of an entry's child entries may not sum up
454 to the entry's size as they normally do.
455</pre>
456<p>However, Massif does not detect and warn about every such occurrence.
457Fortunately, malformed stack traces are rare in practice.</p>
458<p>Returning now to ms_print's output, the final part is similar:</p>
459<pre class="screen">
460--------------------------------------------------------------------------------
461 n time(B) total(B) useful-heap(B) extra-heap(B) stacks(B)
462--------------------------------------------------------------------------------
463 15 21,112 19,096 19,000 96 0
464 16 22,120 18,088 18,000 88 0
465 17 23,128 17,080 17,000 80 0
466 18 24,136 16,072 16,000 72 0
467 19 25,144 15,064 15,000 64 0
468 20 26,152 14,056 14,000 56 0
469 21 27,160 13,048 13,000 48 0
470 22 28,168 12,040 12,000 40 0
471 23 29,176 11,032 11,000 32 0
472 24 30,184 10,024 10,000 24 0
47399.76% (10,000B) (heap allocation functions) malloc/new/new[], --alloc-fns, etc.
474-&gt;79.81% (8,000B) 0x80483C2: g (example.c:5)
475| -&gt;39.90% (4,000B) 0x80483E2: f (example.c:11)
476| | -&gt;39.90% (4,000B) 0x8048431: main (example.c:23)
477| |
478| -&gt;39.90% (4,000B) 0x8048436: main (example.c:25)
479|
480-&gt;19.95% (2,000B) 0x80483DA: f (example.c:10)
481| -&gt;19.95% (2,000B) 0x8048431: main (example.c:23)
482|
483-&gt;00.00% (0B) in 1+ places, all below ms_print's threshold (01.00%)
484</pre>
485<p>The final detailed snapshot shows how the heap looked at termination.
486The 00.00% entry represents the code locations for which memory was
487allocated and then freed (line 20 in this case, the memory for which was
488freed on line 28). However, no code location details are given for this
489entry; by default, Massif only records the details for code locations
490responsible for more than 1% of useful memory bytes, and ms_print likewise
491only prints the details for code locations responsible for more than 1%.
492The entries that do not meet this threshold are aggregated. This avoids
493filling up the output with large numbers of unimportant entries. The
494thresholds can be changed with the
495<code class="option">--threshold</code> option that both Massif and
496ms_print support.</p>
497</div>
498<div class="sect2">
499<div class="titlepage"><div><div><h3 class="title">
500<a name="ms-manual.forkingprograms"></a>9.2.7. Forking Programs</h3></div></div></div>
501<p>If your program forks, the child will inherit all the profiling data that
502has been gathered for the parent.</p>
503<p>If the output file format string (controlled by
504<code class="option">--massif-out-file</code>) does not contain <code class="option">%p</code>, then
505the outputs from the parent and child will be intermingled in a single output
506file, which will almost certainly make it unreadable by ms_print.</p>
507</div>
508<div class="sect2">
509<div class="titlepage"><div><div><h3 class="title">
510<a name="ms-manual.not-measured"></a>9.2.8. Measuring All Memory in a Process</h3></div></div></div>
511<p>
512It is worth emphasising that by default Massif measures only heap memory, i.e.
513memory allocated with
514<code class="function">malloc</code>,
515<code class="function">calloc</code>,
516<code class="function">realloc</code>,
517<code class="function">memalign</code>,
518<code class="function">new</code>,
519<code class="function">new[]</code>,
520and a few other, similar functions. (And it can optionally measure stack
521memory, of course.) This means it does <span class="emphasis"><em>not</em></span> directly
522measure memory allocated with lower-level system calls such as
523<code class="function">mmap</code>,
524<code class="function">mremap</code>, and
525<code class="function">brk</code>.
526</p>
527<p>
528Heap allocation functions such as <code class="function">malloc</code> are built on
529top of these system calls. For example, when needed, an allocator will
530typically call <code class="function">mmap</code> to allocate a large chunk of
531memory, and then hand over pieces of that memory chunk to the client program
532in response to calls to <code class="function">malloc</code> et al. Massif directly
533measures only these higher-level <code class="function">malloc</code> et al calls,
534not the lower-level system calls.
535</p>
536<p>
537Furthermore, a client program may use these lower-level system calls
538directly to allocate memory. By default, Massif does not measure these. Nor
539does it measure the size of code, data and BSS segments. Therefore, the
540numbers reported by Massif may be significantly smaller than those reported by
541tools such as <code class="filename">top</code> that measure a program's total size in
542memory.
543</p>
544<p>
545However, if you wish to measure <span class="emphasis"><em>all</em></span> the memory used by
546your program, you can use the <code class="option">--pages-as-heap=yes</code>. When this
547option is enabled, Massif's normal heap block profiling is replaced by
548lower-level page profiling. Every page allocated via
549<code class="function">mmap</code> and similar system calls is treated as a distinct
550block. This means that code, data and BSS segments are all measured, as they
551are just memory pages. Even the stack is measured, since it is ultimately
552allocated (and extended when necessary) via <code class="function">mmap</code>; for
553this reason <code class="option">--stacks=yes</code> is not allowed in conjunction with
554<code class="option">--pages-as-heap=yes</code>.
555</p>
556<p>
557After <code class="option">--pages-as-heap=yes</code> is used, ms_print's output is
558mostly unchanged. One difference is that the start of each detailed snapshot
559says:
560</p>
561<pre class="screen">
562(page allocation syscalls) mmap/mremap/brk, --alloc-fns, etc.
563</pre>
564<p>instead of the usual</p>:
565
566<pre class="screen">
567(heap allocation functions) malloc/new/new[], --alloc-fns, etc.
568</pre>
569<p>
570The stack traces in the output may be more difficult to read, and interpreting
571them may require some detailed understanding of the lower levels of a program
572like the memory allocators. But for some programs having the full information
573about memory usage can be very useful.
574</p>
575</div>
576<div class="sect2">
577<div class="titlepage"><div><div><h3 class="title">
578<a name="ms-manual.acting"></a>9.2.9. Acting on Massif's Information</h3></div></div></div>
579<p>Massif's information is generally fairly easy to act upon. The
580obvious place to start looking is the peak snapshot.</p>
581<p>It can also be useful to look at the overall shape of the graph, to
582see if memory usage climbs and falls as you expect; spikes in the graph
583might be worth investigating.</p>
584<p>The detailed snapshots can get quite large. It is worth viewing them
585in a very wide window. It's also a good idea to view them with a text
586editor. That makes it easy to scroll up and down while keeping the cursor
587in a particular column, which makes following the allocation chains easier.
588</p>
589</div>
590</div>
591<div class="sect1">
592<div class="titlepage"><div><div><h2 class="title" style="clear: both">
593<a name="ms-manual.options"></a>9.3. Massif Command-line Options</h2></div></div></div>
594<p>Massif-specific command-line options are:</p>
595<div class="variablelist">
596<a name="ms.opts.list"></a><dl class="variablelist">
597<dt>
598<a name="opt.heap"></a><span class="term">
599 <code class="option">--heap=&lt;yes|no&gt; [default: yes] </code>
600 </span>
601</dt>
602<dd><p>Specifies whether heap profiling should be done.</p></dd>
603<dt>
604<a name="opt.heap-admin"></a><span class="term">
605 <code class="option">--heap-admin=&lt;size&gt; [default: 8] </code>
606 </span>
607</dt>
608<dd><p>If heap profiling is enabled, gives the number of administrative
609 bytes per block to use. This should be an estimate of the average,
610 since it may vary. For example, the allocator used by
611 glibc on Linux requires somewhere between 4 to
612 15 bytes per block, depending on various factors. That allocator also
613 requires admin space for freed blocks, but Massif cannot
614 account for this.</p></dd>
615<dt>
616<a name="opt.stacks"></a><span class="term">
617 <code class="option">--stacks=&lt;yes|no&gt; [default: no] </code>
618 </span>
619</dt>
620<dd><p>Specifies whether stack profiling should be done. This option
621 slows Massif down greatly, and so is off by default. Note that Massif
622 assumes that the main stack has size zero at start-up. This is not
623 true, but doing otherwise accurately is difficult. Furthermore,
624 starting at zero better indicates the size of the part of the main
625 stack that a user program actually has control over.</p></dd>
626<dt>
627<a name="opt.pages-as-heap"></a><span class="term">
628 <code class="option">--pages-as-heap=&lt;yes|no&gt; [default: no] </code>
629 </span>
630</dt>
631<dd><p>Tells Massif to profile memory at the page level rather
632 than at the malloc'd block level. See above for details.
633 </p></dd>
634<dt>
635<a name="opt.depth"></a><span class="term">
636 <code class="option">--depth=&lt;number&gt; [default: 30] </code>
637 </span>
638</dt>
639<dd><p>Maximum depth of the allocation trees recorded for detailed
640 snapshots. Increasing it will make Massif run somewhat more slowly,
641 use more memory, and produce bigger output files.</p></dd>
642<dt>
643<a name="opt.alloc-fn"></a><span class="term">
644 <code class="option">--alloc-fn=&lt;name&gt; </code>
645 </span>
646</dt>
647<dd>
648<p>Functions specified with this option will be treated as though
649 they were a heap allocation function such as
650 <code class="function">malloc</code>. This is useful for functions that are
651 wrappers to <code class="function">malloc</code> or <code class="function">new</code>,
652 which can fill up the allocation trees with uninteresting information.
653 This option can be specified multiple times on the command line, to
654 name multiple functions.</p>
655<p>Note that the named function will only be treated this way if it is
656 the top entry in a stack trace, or just below another function treated
657 this way. For example, if you have a function
658 <code class="function">malloc1</code> that wraps <code class="function">malloc</code>,
659 and <code class="function">malloc2</code> that wraps
660 <code class="function">malloc1</code>, just specifying
661 <code class="option">--alloc-fn=malloc2</code> will have no effect. You need to
662 specify <code class="option">--alloc-fn=malloc1</code> as well. This is a little
663 inconvenient, but the reason is that checking for allocation functions
664 is slow, and it saves a lot of time if Massif can stop looking through
665 the stack trace entries as soon as it finds one that doesn't match
666 rather than having to continue through all the entries.</p>
667<p>Note that C++ names are demangled. Note also that overloaded
668 C++ names must be written in full. Single quotes may be necessary to
669 prevent the shell from breaking them up. For example:
670</p>
671<pre class="screen">
672--alloc-fn='operator new(unsigned, std::nothrow_t const&amp;)'
673</pre>
674<p>
675 </p>
676</dd>
677<dt>
678<a name="opt.ignore-fn"></a><span class="term">
679 <code class="option">--ignore-fn=&lt;name&gt; </code>
680 </span>
681</dt>
682<dd>
683<p>Any direct heap allocation (i.e. a call to
684 <code class="function">malloc</code>, <code class="function">new</code>, etc, or a call
685 to a function named by an <code class="option">--alloc-fn</code>
686 option) that occurs in a function specified by this option will be
687 ignored. This is mostly useful for testing purposes. This option can
688 be specified multiple times on the command line, to name multiple
689 functions.
690 </p>
691<p>Any <code class="function">realloc</code> of an ignored block will
692 also be ignored, even if the <code class="function">realloc</code> call does
693 not occur in an ignored function. This avoids the possibility of
694 negative heap sizes if ignored blocks are shrunk with
695 <code class="function">realloc</code>.
696 </p>
697<p>The rules for writing C++ function names are the same as
698 for <code class="option">--alloc-fn</code> above.
699 </p>
700</dd>
701<dt>
702<a name="opt.threshold"></a><span class="term">
703 <code class="option">--threshold=&lt;m.n&gt; [default: 1.0] </code>
704 </span>
705</dt>
706<dd><p>The significance threshold for heap allocations, as a
707 percentage of total memory size. Allocation tree entries that account
708 for less than this will be aggregated. Note that this should be
709 specified in tandem with ms_print's option of the same name.</p></dd>
710<dt>
711<a name="opt.peak-inaccuracy"></a><span class="term">
712 <code class="option">--peak-inaccuracy=&lt;m.n&gt; [default: 1.0] </code>
713 </span>
714</dt>
715<dd><p>Massif does not necessarily record the actual global memory
716 allocation peak; by default it records a peak only when the global
717 memory allocation size exceeds the previous peak by at least 1.0%.
718 This is because there can be many local allocation peaks along the way,
719 and doing a detailed snapshot for every one would be expensive and
720 wasteful, as all but one of them will be later discarded. This
721 inaccuracy can be changed (even to 0.0%) via this option, but Massif
722 will run drastically slower as the number approaches zero.</p></dd>
723<dt>
724<a name="opt.time-unit"></a><span class="term">
725 <code class="option">--time-unit=&lt;i|ms|B&gt; [default: i] </code>
726 </span>
727</dt>
728<dd><p>The time unit used for the profiling. There are three
729 possibilities: instructions executed (i), which is good for most
730 cases; real (wallclock) time (ms, i.e. milliseconds), which is
731 sometimes useful; and bytes allocated/deallocated on the heap and/or
732 stack (B), which is useful for very short-run programs, and for
733 testing purposes, because it is the most reproducible across different
734 machines.</p></dd>
735<dt>
736<a name="opt.detailed-freq"></a><span class="term">
737 <code class="option">--detailed-freq=&lt;n&gt; [default: 10] </code>
738 </span>
739</dt>
740<dd><p>Frequency of detailed snapshots. With
741 <code class="option">--detailed-freq=1</code>, every snapshot is
742 detailed.</p></dd>
743<dt>
744<a name="opt.max-snapshots"></a><span class="term">
745 <code class="option">--max-snapshots=&lt;n&gt; [default: 100] </code>
746 </span>
747</dt>
748<dd><p>The maximum number of snapshots recorded. If set to N, for all
749 programs except very short-running ones, the final number of snapshots
750 will be between N/2 and N.</p></dd>
751<dt>
752<a name="opt.massif-out-file"></a><span class="term">
753 <code class="option">--massif-out-file=&lt;file&gt; [default: massif.out.%p] </code>
754 </span>
755</dt>
756<dd><p>Write the profile data to <code class="computeroutput">file</code>
757 rather than to the default output file,
758 <code class="computeroutput">massif.out.&lt;pid&gt;</code>. The
759 <code class="option">%p</code> and <code class="option">%q</code> format specifiers can be
760 used to embed the process ID and/or the contents of an environment
761 variable in the name, as is the case for the core option
762 <code class="option"><a class="xref" href="manual-core.html#opt.log-file">--log-file</a></code>.
763 </p></dd>
764</dl>
765</div>
766</div>
767<div class="sect1">
768<div class="titlepage"><div><div><h2 class="title" style="clear: both">
769<a name="ms-manual.monitor-commands"></a>9.4. Massif Monitor Commands</h2></div></div></div>
770<p>The Massif tool provides monitor commands handled by the Valgrind
771gdbserver (see <a class="xref" href="manual-core-adv.html#manual-core-adv.gdbserver-commandhandling" title="3.2.5. Monitor command handling by the Valgrind gdbserver">Monitor command handling by the Valgrind gdbserver</a>).
772</p>
773<div class="itemizedlist"><ul class="itemizedlist" style="list-style-type: disc; ">
774<li class="listitem"><p><code class="varname">snapshot [&lt;filename&gt;]</code> requests
775 to take a snapshot and save it in the given &lt;filename&gt;
776 (default massif.vgdb.out).
777 </p></li>
778<li class="listitem"><p><code class="varname">detailed_snapshot [&lt;filename&gt;]</code>
779 requests to take a detailed snapshot and save it in the given
780 &lt;filename&gt; (default massif.vgdb.out).
781 </p></li>
782<li class="listitem"><p><code class="varname">all_snapshots [&lt;filename&gt;]</code>
783 requests to take all captured snapshots so far and save them in the given
784 &lt;filename&gt; (default massif.vgdb.out).
785 </p></li>
786</ul></div>
787</div>
788<div class="sect1">
789<div class="titlepage"><div><div><h2 class="title" style="clear: both">
790<a name="ms-manual.clientreqs"></a>9.5. Massif Client Requests</h2></div></div></div>
791<p>Massif does not have a <code class="filename">massif.h</code> file, but it does
792implement two of the core client requests:
793<code class="function">VALGRIND_MALLOCLIKE_BLOCK</code> and
794<code class="function">VALGRIND_FREELIKE_BLOCK</code>; they are described in
795<a class="xref" href="manual-core-adv.html#manual-core-adv.clientreq" title="3.1. The Client Request mechanism">The Client Request mechanism</a>.
796</p>
797</div>
798<div class="sect1">
799<div class="titlepage"><div><div><h2 class="title" style="clear: both">
800<a name="ms-manual.ms_print-options"></a>9.6. ms_print Command-line Options</h2></div></div></div>
801<p>ms_print's options are:</p>
802<div class="variablelist">
803<a name="ms_print.opts.list"></a><dl class="variablelist">
804<dt><span class="term">
805 <code class="option">-h --help </code>
806 </span></dt>
807<dd><p>Show the help message.</p></dd>
808<dt><span class="term">
809 <code class="option">--version </code>
810 </span></dt>
811<dd><p>Show the version number.</p></dd>
812<dt><span class="term">
813 <code class="option">--threshold=&lt;m.n&gt; [default: 1.0] </code>
814 </span></dt>
815<dd><p>Same as Massif's <code class="option">--threshold</code> option, but
816 applied after profiling rather than during.</p></dd>
817<dt><span class="term">
818 <code class="option">--x=&lt;4..1000&gt; [default: 72]</code>
819 </span></dt>
820<dd><p>Width of the graph, in columns.</p></dd>
821<dt><span class="term">
822 <code class="option">--y=&lt;4..1000&gt; [default: 20] </code>
823 </span></dt>
824<dd><p>Height of the graph, in rows.</p></dd>
825</dl>
826</div>
827</div>
828<div class="sect1">
829<div class="titlepage"><div><div><h2 class="title" style="clear: both">
830<a name="ms-manual.fileformat"></a>9.7. Massif's Output File Format</h2></div></div></div>
831<p>Massif's file format is plain text (i.e. not binary) and deliberately
832easy to read for both humans and machines. Nonetheless, the exact format
833is not described here. This is because the format is currently very
834Massif-specific. In the future we hope to make the format more general, and
835thus suitable for possible use with other tools. Once this has been done,
836the format will be documented here.</p>
837</div>
838</div>
839<div>
840<br><table class="nav" width="100%" cellspacing="3" cellpadding="2" border="0" summary="Navigation footer">
841<tr>
842<td rowspan="2" width="40%" align="left">
843<a accesskey="p" href="drd-manual.html">&lt;&lt; 8. DRD: a thread error detector</a> </td>
844<td width="20%" align="center"><a accesskey="u" href="manual.html">Up</a></td>
845<td rowspan="2" width="40%" align="right"> <a accesskey="n" href="dh-manual.html">10. DHAT: a dynamic heap analysis tool &gt;&gt;</a>
846</td>
847</tr>
848<tr><td width="20%" align="center"><a accesskey="h" href="index.html">Home</a></td></tr>
849</table>
850</div>
851</body>
852</html>