1VALGRIND(1) valgrind VALGRIND(1)
2
6 valgrind - a suite of tools for debugging and profiling programs
7
9 valgrind [valgrind-options] [your-program] [your-program-options]
10
12 Valgrind is a flexible program for debugging and profiling Linux
13 executables. It consists of a core, which provides a synthetic CPU in
14 software, and a series of debugging and profiling tools. The
15 architecture is modular, so that new tools can be created easily and
16 without disturbing the existing structure.
17 Some of the options described below work with all Valgrind tools, and
19 some only work with a few or one. The section MEMCHECK OPTIONS and
20 those below it describe tool-specific options.
21 This manual page covers only basic usage and options. For more
23 comprehensive information, please see the HTML documentation on your
24 system: $INSTALL/share/doc/valgrind/html/index.html, or online:
25 http://www.valgrind.org/docs/manual/index.html.
26
28 The single most important option.
29 --tool=<toolname> [default: memcheck]
31 Run the Valgrind tool called toolname, e.g. memcheck, cachegrind,
32 callgrind, helgrind, drd, massif, dhat, lackey, none, exp-bbv, etc.
33
35 These options work with all tools.
36 -h --help
38 Show help for all options, both for the core and for the selected
39 tool. If the option is repeated it is equivalent to giving
40 --help-debug.
41 --help-debug
43 Same as --help, but also lists debugging options which usually are
44 only of use to Valgrind's developers.
45 --version
47 Show the version number of the Valgrind core. Tools can have their
48 own version numbers. There is a scheme in place to ensure that
49 tools only execute when the core version is one they are known to
50 work with. This was done to minimise the chances of strange
51 problems arising from tool-vs-core version incompatibilities.
52 -q, --quiet
54 Run silently, and only print error messages. Useful if you are
55 running regression tests or have some other automated test
56 machinery.
57 -v, --verbose
59 Be more verbose. Gives extra information on various aspects of your
60 program, such as: the shared objects loaded, the suppressions used,
61 the progress of the instrumentation and execution engines, and
62 warnings about unusual behaviour. Repeating the option increases
63 the verbosity level.
64 --trace-children=<yes|no> [default: no]
66 When enabled, Valgrind will trace into sub-processes initiated via
67 the exec system call. This is necessary for multi-process programs.
68 Note that Valgrind does trace into the child of a fork (it would be
70 difficult not to, since fork makes an identical copy of a process),
71 so this option is arguably badly named. However, most children of
72 fork calls immediately call exec anyway.
73 --trace-children-skip=patt1,patt2,...
75 This option only has an effect when --trace-children=yes is
76 specified. It allows for some children to be skipped. The option
77 takes a comma separated list of patterns for the names of child
78 executables that Valgrind should not trace into. Patterns may
79 include the metacharacters ? and *, which have the usual meaning.
80 This can be useful for pruning uninteresting branches from a tree
82 of processes being run on Valgrind. But you should be careful when
83 using it. When Valgrind skips tracing into an executable, it
84 doesn't just skip tracing that executable, it also skips tracing
85 any of that executable's child processes. In other words, the flag
86 doesn't merely cause tracing to stop at the specified executables
87 -- it skips tracing of entire process subtrees rooted at any of the
88 specified executables.
89 --trace-children-skip-by-arg=patt1,patt2,...
91 This is the same as --trace-children-skip, with one difference: the
92 decision as to whether to trace into a child process is made by
93 examining the arguments to the child process, rather than the name
94 of its executable.
95 --child-silent-after-fork=<yes|no> [default: no]
97 When enabled, Valgrind will not show any debugging or logging
98 output for the child process resulting from a fork call. This can
99 make the output less confusing (although more misleading) when
100 dealing with processes that create children. It is particularly
101 useful in conjunction with --trace-children=. Use of this option is
102 also strongly recommended if you are requesting XML output
103 (--xml=yes), since otherwise the XML from child and parent may
104 become mixed up, which usually makes it useless.
105 --vgdb=<no|yes|full> [default: yes]
107 Valgrind will provide "gdbserver" functionality when --vgdb=yes or
108 --vgdb=full is specified. This allows an external GNU GDB debugger
109 to control and debug your program when it runs on Valgrind.
110 --vgdb=full incurs significant performance overheads, but provides
111 more precise breakpoints and watchpoints. See Debugging your
112 program using Valgrind's gdbserver and GDB for a detailed
113 description.
114 If the embedded gdbserver is enabled but no gdb is currently being
116 used, the vgdb command line utility can send "monitor commands" to
117 Valgrind from a shell. The Valgrind core provides a set of Valgrind
118 monitor commands. A tool can optionally provide tool specific
119 monitor commands, which are documented in the tool specific
120 chapter.
121 --vgdb-error=<number> [default: 999999999]
123 Use this option when the Valgrind gdbserver is enabled with
124 --vgdb=yes or --vgdb=full. Tools that report errors will wait for
125 "number" errors to be reported before freezing the program and
126 waiting for you to connect with GDB. It follows that a value of
127 zero will cause the gdbserver to be started before your program is
128 executed. This is typically used to insert GDB breakpoints before
129 execution, and also works with tools that do not report errors,
130 such as Massif.
131 --vgdb-stop-at=<set> [default: none]
133 Use this option when the Valgrind gdbserver is enabled with
134 --vgdb=yes or --vgdb=full. The Valgrind gdbserver will be invoked
135 for each error after --vgdb-error have been reported. You can
136 additionally ask the Valgrind gdbserver to be invoked for other
137 events, specified in one of the following ways:
138 • a comma separated list of one or more of startup exit
140 valgrindabexit.
141 The values startup exit valgrindabexit respectively indicate to
143 invoke gdbserver before your program is executed, after the
144 last instruction of your program, on Valgrind abnormal exit
145 (e.g. internal error, out of memory, ...).
146 Note: startup and --vgdb-error=0 will both cause Valgrind
148 gdbserver to be invoked before your program is executed. The
149 --vgdb-error=0 will in addition cause your program to stop on
150 all subsequent errors.
151 • all to specify the complete set. It is equivalent to
153 --vgdb-stop-at=startup,exit,valgrindabexit.
154 • none for the empty set.
156 --track-fds=<yes|no|all> [default: no]
158 When enabled, Valgrind will print out a list of open file
159 descriptors on exit or on request, via the gdbserver monitor
160 command v.info open_fds. Along with each file descriptor is printed
161 a stack backtrace of where the file was opened and any details
162 relating to the file descriptor such as the file name or socket
163 details. Use all to include reporting on stdin, stdout and stderr.
164 --time-stamp=<yes|no> [default: no]
166 When enabled, each message is preceded with an indication of the
167 elapsed wallclock time since startup, expressed as days, hours,
168 minutes, seconds and milliseconds.
169 --log-fd=<number> [default: 2, stderr]
171 Specifies that Valgrind should send all of its messages to the
172 specified file descriptor. The default, 2, is the standard error
173 channel (stderr). Note that this may interfere with the client's
174 own use of stderr, as Valgrind's output will be interleaved with
175 any output that the client sends to stderr.
176 --log-file=<filename>
178 Specifies that Valgrind should send all of its messages to the
179 specified file. If the file name is empty, it causes an abort.
180 There are three special format specifiers that can be used in the
181 file name.
182 %p is replaced with the current process ID. This is very useful for
184 program that invoke multiple processes. WARNING: If you use
185 --trace-children=yes and your program invokes multiple processes OR
186 your program forks without calling exec afterwards, and you don't
187 use this specifier (or the %q specifier below), the Valgrind output
188 from all those processes will go into one file, possibly jumbled
189 up, and possibly incomplete. Note: If the program forks and calls
190 exec afterwards, Valgrind output of the child from the period
191 between fork and exec will be lost. Fortunately this gap is really
192 tiny for most programs; and modern programs use posix_spawn anyway.
193 %n is replaced with a file sequence number unique for this process.
195 This is useful for processes that produces several files from the
196 same filename template.
197 %q{FOO} is replaced with the contents of the environment variable
199 FOO. If the {FOO} part is malformed, it causes an abort. This
200 specifier is rarely needed, but very useful in certain
201 circumstances (eg. when running MPI programs). The idea is that you
202 specify a variable which will be set differently for each process
203 in the job, for example BPROC_RANK or whatever is applicable in
204 your MPI setup. If the named environment variable is not set, it
205 causes an abort. Note that in some shells, the { and } characters
206 may need to be escaped with a backslash.
207 %% is replaced with %.
209 If an % is followed by any other character, it causes an abort.
211 If the file name specifies a relative file name, it is put in the
213 program's initial working directory: this is the current directory
214 when the program started its execution after the fork or after the
215 exec. If it specifies an absolute file name (ie. starts with '/')
216 then it is put there.
217 --log-socket=<ip-address:port-number>
219 Specifies that Valgrind should send all of its messages to the
220 specified port at the specified IP address. The port may be
221 omitted, in which case port 1500 is used. If a connection cannot be
222 made to the specified socket, Valgrind falls back to writing output
223 to the standard error (stderr). This option is intended to be used
224 in conjunction with the valgrind-listener program. For further
225 details, see the commentary in the manual.
226
228 These options are used by all tools that can report errors, e.g.
229 Memcheck, but not Cachegrind.
230 --xml=<yes|no> [default: no]
232 When enabled, the important parts of the output (e.g. tool error
233 messages) will be in XML format rather than plain text.
234 Furthermore, the XML output will be sent to a different output
235 channel than the plain text output. Therefore, you also must use
236 one of --xml-fd, --xml-file or --xml-socket to specify where the
237 XML is to be sent.
238 Less important messages will still be printed in plain text, but
240 because the XML output and plain text output are sent to different
241 output channels (the destination of the plain text output is still
242 controlled by --log-fd, --log-file and --log-socket) this should
243 not cause problems.
244 This option is aimed at making life easier for tools that consume
246 Valgrind's output as input, such as GUI front ends. Currently this
247 option works with Memcheck, Helgrind and DRD. The output format is
248 specified in the file docs/internals/xml-output-protocol4.txt in
249 the source tree for Valgrind 3.5.0 or later.
250 The recommended options for a GUI to pass, when requesting XML
252 output, are: --xml=yes to enable XML output, --xml-file to send the
253 XML output to a (presumably GUI-selected) file, --log-file to send
254 the plain text output to a second GUI-selected file,
255 --child-silent-after-fork=yes, and -q to restrict the plain text
256 output to critical error messages created by Valgrind itself. For
257 example, failure to read a specified suppressions file counts as a
258 critical error message. In this way, for a successful run the text
259 output file will be empty. But if it isn't empty, then it will
260 contain important information which the GUI user should be made
261 aware of.
262 --xml-fd=<number> [default: -1, disabled]
264 Specifies that Valgrind should send its XML output to the specified
265 file descriptor. It must be used in conjunction with --xml=yes.
266 --xml-file=<filename>
268 Specifies that Valgrind should send its XML output to the specified
269 file. It must be used in conjunction with --xml=yes. Any %p or %q
270 sequences appearing in the filename are expanded in exactly the
271 same way as they are for --log-file. See the description of --log-
272 file for details.
273 --xml-socket=<ip-address:port-number>
275 Specifies that Valgrind should send its XML output the specified
276 port at the specified IP address. It must be used in conjunction
277 with --xml=yes. The form of the argument is the same as that used
278 by --log-socket. See the description of --log-socket for further
279 details.
280 --xml-user-comment=<string>
282 Embeds an extra user comment string at the start of the XML output.
283 Only works when --xml=yes is specified; ignored otherwise.
284 --demangle=<yes|no> [default: yes]
286 Enable/disable automatic demangling (decoding) of C++ names.
287 Enabled by default. When enabled, Valgrind will attempt to
288 translate encoded C++ names back to something approaching the
289 original. The demangler handles symbols mangled by g++ versions
290 2.X, 3.X and 4.X.
291 An important fact about demangling is that function names mentioned
293 in suppressions files should be in their mangled form. Valgrind
294 does not demangle function names when searching for applicable
295 suppressions, because to do otherwise would make suppression file
296 contents dependent on the state of Valgrind's demangling machinery,
297 and also slow down suppression matching.
298 --num-callers=<number> [default: 12]
300 Specifies the maximum number of entries shown in stack traces that
301 identify program locations. Note that errors are commoned up using
302 only the top four function locations (the place in the current
303 function, and that of its three immediate callers). So this doesn't
304 affect the total number of errors reported.
305 The maximum value for this is 500. Note that higher settings will
307 make Valgrind run a bit more slowly and take a bit more memory, but
308 can be useful when working with programs with deeply-nested call
309 chains.
310 --unw-stack-scan-thresh=<number> [default: 0] ,
312 --unw-stack-scan-frames=<number> [default: 5]
313 Stack-scanning support is available only on ARM targets.
314 These flags enable and control stack unwinding by stack scanning.
316 When the normal stack unwinding mechanisms -- usage of Dwarf CFI
317 records, and frame-pointer following -- fail, stack scanning may be
318 able to recover a stack trace.
319 Note that stack scanning is an imprecise, heuristic mechanism that
321 may give very misleading results, or none at all. It should be used
322 only in emergencies, when normal unwinding fails, and it is
323 important to nevertheless have stack traces.
324 Stack scanning is a simple technique: the unwinder reads words from
326 the stack, and tries to guess which of them might be return
327 addresses, by checking to see if they point just after ARM or Thumb
328 call instructions. If so, the word is added to the backtrace.
329 The main danger occurs when a function call returns, leaving its
331 return address exposed, and a new function is called, but the new
332 function does not overwrite the old address. The result of this is
333 that the backtrace may contain entries for functions which have
334 already returned, and so be very confusing.
335 A second limitation of this implementation is that it will scan
337 only the page (4KB, normally) containing the starting stack
338 pointer. If the stack frames are large, this may result in only a
339 few (or not even any) being present in the trace. Also, if you are
340 unlucky and have an initial stack pointer near the end of its
341 containing page, the scan may miss all interesting frames.
342 By default stack scanning is disabled. The normal use case is to
344 ask for it when a stack trace would otherwise be very short. So, to
345 enable it, use --unw-stack-scan-thresh=number. This requests
346 Valgrind to try using stack scanning to "extend" stack traces which
347 contain fewer than number frames.
348 If stack scanning does take place, it will only generate at most
350 the number of frames specified by --unw-stack-scan-frames.
351 Typically, stack scanning generates so many garbage entries that
352 this value is set to a low value (5) by default. In no case will a
353 stack trace larger than the value specified by --num-callers be
354 created.
355 --error-limit=<yes|no> [default: yes]
357 When enabled, Valgrind stops reporting errors after 10,000,000 in
358 total, or 1,000 different ones, have been seen. This is to stop the
359 error tracking machinery from becoming a huge performance overhead
360 in programs with many errors.
361 --error-exitcode=<number> [default: 0]
363 Specifies an alternative exit code to return if Valgrind reported
364 any errors in the run. When set to the default value (zero), the
365 return value from Valgrind will always be the return value of the
366 process being simulated. When set to a nonzero value, that value is
367 returned instead, if Valgrind detects any errors. This is useful
368 for using Valgrind as part of an automated test suite, since it
369 makes it easy to detect test cases for which Valgrind has reported
370 errors, just by inspecting return codes.
371 --exit-on-first-error=<yes|no> [default: no]
373 If this option is enabled, Valgrind exits on the first error. A
374 nonzero exit value must be defined using --error-exitcode option.
375 Useful if you are running regression tests or have some other
376 automated test machinery.
377 --error-markers=<begin>,<end> [default: none]
379 When errors are output as plain text (i.e. XML not used),
380 --error-markers instructs to output a line containing the begin
381 (end) string before (after) each error.
382 Such marker lines facilitate searching for errors and/or extracting
384 errors in an output file that contain valgrind errors mixed with
385 the program output.
386 Note that empty markers are accepted. So, only using a begin (or an
388 end) marker is possible.
389 --show-error-list=no|yes [default: no]
391 If this option is enabled, for tools that report errors, valgrind
392 will show the list of detected errors and the list of used
393 suppressions at exit.
394 Note that at verbosity 2 and above, valgrind automatically shows
396 the list of detected errors and the list of used suppressions at
397 exit, unless --show-error-list=no is selected.
398 -s
400 Specifying -s is equivalent to --show-error-list=yes.
401 --sigill-diagnostics=<yes|no> [default: yes]
403 Enable/disable printing of illegal instruction diagnostics. Enabled
404 by default, but defaults to disabled when --quiet is given. The
405 default can always be explicitly overridden by giving this option.
406 When enabled, a warning message will be printed, along with some
408 diagnostics, whenever an instruction is encountered that Valgrind
409 cannot decode or translate, before the program is given a SIGILL
410 signal. Often an illegal instruction indicates a bug in the program
411 or missing support for the particular instruction in Valgrind. But
412 some programs do deliberately try to execute an instruction that
413 might be missing and trap the SIGILL signal to detect processor
414 features. Using this flag makes it possible to avoid the diagnostic
415 output that you would otherwise get in such cases.
416 --keep-debuginfo=<yes|no> [default: no]
418 When enabled, keep ("archive") symbols and all other debuginfo for
419 unloaded code. This allows saved stack traces to include file/line
420 info for code that has been dlclose'd (or similar). Be careful with
421 this, since it can lead to unbounded memory use for programs which
422 repeatedly load and unload shared objects.
423 Some tools and some functionalities have only limited support for
425 archived debug info. Memcheck fully supports it. Generally, tools
426 that report errors can use archived debug info to show the error
427 stack traces. The known limitations are: Helgrind's past access
428 stack trace of a race condition is does not use archived debug
429 info. Massif (and more generally the xtree Massif output format)
430 does not make use of archived debug info. Only Memcheck has been
431 (somewhat) tested with --keep-debuginfo=yes, so other tools may
432 have unknown limitations.
433 --show-below-main=<yes|no> [default: no]
435 By default, stack traces for errors do not show any functions that
436 appear beneath main because most of the time it's uninteresting C
437 library stuff and/or gobbledygook. Alternatively, if main is not
438 present in the stack trace, stack traces will not show any
439 functions below main-like functions such as glibc's
440 __libc_start_main. Furthermore, if main-like functions are present
441 in the trace, they are normalised as (below main), in order to make
442 the output more deterministic.
443 If this option is enabled, all stack trace entries will be shown
445 and main-like functions will not be normalised.
446 --fullpath-after=<string> [default: don't show source paths]
448 By default Valgrind only shows the filenames in stack traces, but
449 not full paths to source files. When using Valgrind in large
450 projects where the sources reside in multiple different
451 directories, this can be inconvenient. --fullpath-after provides a
452 flexible solution to this problem. When this option is present, the
453 path to each source file is shown, with the following all-important
454 caveat: if string is found in the path, then the path up to and
455 including string is omitted, else the path is shown unmodified.
456 Note that string is not required to be a prefix of the path.
457 For example, consider a file named
459 /home/janedoe/blah/src/foo/bar/xyzzy.c. Specifying
460 --fullpath-after=/home/janedoe/blah/src/ will cause Valgrind to
461 show the name as foo/bar/xyzzy.c.
462 Because the string is not required to be a prefix,
464 --fullpath-after=src/ will produce the same output. This is useful
465 when the path contains arbitrary machine-generated characters. For
466 example, the path /my/build/dir/C32A1B47/blah/src/foo/xyzzy can be
467 pruned to foo/xyzzy using --fullpath-after=/blah/src/.
468 If you simply want to see the full path, just specify an empty
470 string: --fullpath-after=. This isn't a special case, merely a
471 logical consequence of the above rules.
472 Finally, you can use --fullpath-after multiple times. Any
474 appearance of it causes Valgrind to switch to producing full paths
475 and applying the above filtering rule. Each produced path is
476 compared against all the --fullpath-after-specified strings, in the
477 order specified. The first string to match causes the path to be
478 truncated as described above. If none match, the full path is
479 shown. This facilitates chopping off prefixes when the sources are
480 drawn from a number of unrelated directories.
481 --extra-debuginfo-path=<path> [default: undefined and unused]
483 By default Valgrind searches in several well-known paths for debug
484 objects, such as /usr/lib/debug/.
485 However, there may be scenarios where you may wish to put debug
487 objects at an arbitrary location, such as external storage when
488 running Valgrind on a mobile device with limited local storage.
489 Another example might be a situation where you do not have
490 permission to install debug object packages on the system where you
491 are running Valgrind.
492 In these scenarios, you may provide an absolute path as an extra,
494 final place for Valgrind to search for debug objects by specifying
495 --extra-debuginfo-path=/path/to/debug/objects. The given path will
496 be prepended to the absolute path name of the searched-for object.
497 For example, if Valgrind is looking for the debuginfo for
498 /w/x/y/zz.so and --extra-debuginfo-path=/a/b/c is specified, it
499 will look for a debug object at /a/b/c/w/x/y/zz.so.
500 This flag should only be specified once. If it is specified
502 multiple times, only the last instance is honoured.
503 --debuginfo-server=ipaddr:port [default: undefined and unused]
505 This is a new, experimental, feature introduced in version 3.9.0.
506 In some scenarios it may be convenient to read debuginfo from
508 objects stored on a different machine. With this flag, Valgrind
509 will query a debuginfo server running on ipaddr and listening on
510 port port, if it cannot find the debuginfo object in the local
511 filesystem.
512 The debuginfo server must accept TCP connections on port port. The
514 debuginfo server is contained in the source file
515 auxprogs/valgrind-di-server.c. It will only serve from the
516 directory it is started in. port defaults to 1500 in both client
517 and server if not specified.
518 If Valgrind looks for the debuginfo for /w/x/y/zz.so by using the
520 debuginfo server, it will strip the pathname components and merely
521 request zz.so on the server. That in turn will look only in its
522 current working directory for a matching debuginfo object.
523 The debuginfo data is transmitted in small fragments (8 KB) as
525 requested by Valgrind. Each block is compressed using LZO to reduce
526 transmission time. The implementation has been tuned for best
527 performance over a single-stage 802.11g (WiFi) network link.
528 Note that checks for matching primary vs debug objects, using GNU
530 debuglink CRC scheme, are performed even when using the debuginfo
531 server. To disable such checking, you need to also specify
532 --allow-mismatched-debuginfo=yes.
533 By default the Valgrind build system will build valgrind-di-server
535 for the target platform, which is almost certainly not what you
536 want. So far we have been unable to find out how to get
537 automake/autoconf to build it for the build platform. If you want
538 to use it, you will have to recompile it by hand using the command
539 shown at the top of auxprogs/valgrind-di-server.c.
540 Valgrind can also download debuginfo via debuginfod. See the
542 DEBUGINFOD section for more information.
543 --allow-mismatched-debuginfo=no|yes [no]
545 When reading debuginfo from separate debuginfo objects, Valgrind
546 will by default check that the main and debuginfo objects match,
547 using the GNU debuglink mechanism. This guarantees that it does not
548 read debuginfo from out of date debuginfo objects, and also ensures
549 that Valgrind can't crash as a result of mismatches.
550 This check can be overridden using
552 --allow-mismatched-debuginfo=yes. This may be useful when the
553 debuginfo and main objects have not been split in the proper way.
554 Be careful when using this, though: it disables all consistency
555 checking, and Valgrind has been observed to crash when the main and
556 debuginfo objects don't match.
557 --suppressions=<filename> [default: $PREFIX/lib/valgrind/default.supp]
559 Specifies an extra file from which to read descriptions of errors
560 to suppress. You may use up to 100 extra suppression files.
561 --gen-suppressions=<yes|no|all> [default: no]
563 When set to yes, Valgrind will pause after every error shown and
564 print the line:
565 ---- Print suppression ? --- [Return/N/n/Y/y/C/c] ----
567 Pressing Ret, or N Ret or n Ret, causes Valgrind continue execution
569 without printing a suppression for this error.
570 Pressing Y Ret or y Ret causes Valgrind to write a suppression for
572 this error. You can then cut and paste it into a suppression file
573 if you don't want to hear about the error in the future.
574 When set to all, Valgrind will print a suppression for every
576 reported error, without querying the user.
577 This option is particularly useful with C++ programs, as it prints
579 out the suppressions with mangled names, as required.
580 Note that the suppressions printed are as specific as possible. You
582 may want to common up similar ones, by adding wildcards to function
583 names, and by using frame-level wildcards. The wildcarding
584 facilities are powerful yet flexible, and with a bit of careful
585 editing, you may be able to suppress a whole family of related
586 errors with only a few suppressions.
587 Sometimes two different errors are suppressed by the same
589 suppression, in which case Valgrind will output the suppression
590 more than once, but you only need to have one copy in your
591 suppression file (but having more than one won't cause problems).
592 Also, the suppression name is given as <insert a suppression name
593 here>; the name doesn't really matter, it's only used with the -v
594 option which prints out all used suppression records.
595 --input-fd=<number> [default: 0, stdin]
597 When using --gen-suppressions=yes, Valgrind will stop so as to read
598 keyboard input from you when each error occurs. By default it reads
599 from the standard input (stdin), which is problematic for programs
600 which close stdin. This option allows you to specify an alternative
601 file descriptor from which to read input.
602 --dsymutil=no|yes [yes]
604 This option is only relevant when running Valgrind on Mac OS X.
605 Mac OS X uses a deferred debug information (debuginfo) linking
607 scheme. When object files containing debuginfo are linked into a
608 .dylib or an executable, the debuginfo is not copied into the final
609 file. Instead, the debuginfo must be linked manually by running
610 dsymutil, a system-provided utility, on the executable or .dylib.
611 The resulting combined debuginfo is placed in a directory alongside
612 the executable or .dylib, but with the extension .dSYM.
613 With --dsymutil=no, Valgrind will detect cases where the .dSYM
615 directory is either missing, or is present but does not appear to
616 match the associated executable or .dylib, most likely because it
617 is out of date. In these cases, Valgrind will print a warning
618 message but take no further action.
619 With --dsymutil=yes, Valgrind will, in such cases, automatically
621 run dsymutil as necessary to bring the debuginfo up to date. For
622 all practical purposes, if you always use --dsymutil=yes, then
623 there is never any need to run dsymutil manually or as part of your
624 applications's build system, since Valgrind will run it as
625 necessary.
626 Valgrind will not attempt to run dsymutil on any executable or
628 library in /usr/, /bin/, /sbin/, /opt/, /sw/, /System/, /Library/
629 or /Applications/ since dsymutil will always fail in such
630 situations. It fails both because the debuginfo for such
631 pre-installed system components is not available anywhere, and also
632 because it would require write privileges in those directories.
633 Be careful when using --dsymutil=yes, since it will cause
635 pre-existing .dSYM directories to be silently deleted and
636 re-created. Also note that dsymutil is quite slow, sometimes
637 excessively so.
638 --max-stackframe=<number> [default: 2000000]
640 The maximum size of a stack frame. If the stack pointer moves by
641 more than this amount then Valgrind will assume that the program is
642 switching to a different stack.
643 You may need to use this option if your program has large
645 stack-allocated arrays. Valgrind keeps track of your program's
646 stack pointer. If it changes by more than the threshold amount,
647 Valgrind assumes your program is switching to a different stack,
648 and Memcheck behaves differently than it would for a stack pointer
649 change smaller than the threshold. Usually this heuristic works
650 well. However, if your program allocates large structures on the
651 stack, this heuristic will be fooled, and Memcheck will
652 subsequently report large numbers of invalid stack accesses. This
653 option allows you to change the threshold to a different value.
654 You should only consider use of this option if Valgrind's debug
656 output directs you to do so. In that case it will tell you the new
657 threshold you should specify.
658 In general, allocating large structures on the stack is a bad idea,
660 because you can easily run out of stack space, especially on
661 systems with limited memory or which expect to support large
662 numbers of threads each with a small stack, and also because the
663 error checking performed by Memcheck is more effective for
664 heap-allocated data than for stack-allocated data. If you have to
665 use this option, you may wish to consider rewriting your code to
666 allocate on the heap rather than on the stack.
667 --main-stacksize=<number> [default: use current 'ulimit' value]
669 Specifies the size of the main thread's stack.
670 To simplify its memory management, Valgrind reserves all required
672 space for the main thread's stack at startup. That means it needs
673 to know the required stack size at startup.
674 By default, Valgrind uses the current "ulimit" value for the stack
676 size, or 16 MB, whichever is lower. In many cases this gives a
677 stack size in the range 8 to 16 MB, which almost never overflows
678 for most applications.
679 If you need a larger total stack size, use --main-stacksize to
681 specify it. Only set it as high as you need, since reserving far
682 more space than you need (that is, hundreds of megabytes more than
683 you need) constrains Valgrind's memory allocators and may reduce
684 the total amount of memory that Valgrind can use. This is only
685 really of significance on 32-bit machines.
686 On Linux, you may request a stack of size up to 2GB. Valgrind will
688 stop with a diagnostic message if the stack cannot be allocated.
689 --main-stacksize only affects the stack size for the program's
691 initial thread. It has no bearing on the size of thread stacks, as
692 Valgrind does not allocate those.
693 You may need to use both --main-stacksize and --max-stackframe
695 together. It is important to understand that --main-stacksize sets
696 the maximum total stack size, whilst --max-stackframe specifies the
697 largest size of any one stack frame. You will have to work out the
698 --main-stacksize value for yourself (usually, if your applications
699 segfaults). But Valgrind will tell you the needed --max-stackframe
700 size, if necessary.
701 As discussed further in the description of --max-stackframe, a
703 requirement for a large stack is a sign of potential portability
704 problems. You are best advised to place all large data in
705 heap-allocated memory.
706 --max-threads=<number> [default: 500]
708 By default, Valgrind can handle to up to 500 threads. Occasionally,
709 that number is too small. Use this option to provide a different
710 limit. E.g. --max-threads=3000.
711
713 For tools that use their own version of malloc (e.g. Memcheck, Massif,
714 Helgrind, DRD), the following options apply.
715 --alignment=<number> [default: 8 or 16, depending on the platform]
717 By default Valgrind's malloc, realloc, etc, return a block whose
718 starting address is 8-byte aligned or 16-byte aligned (the value
719 depends on the platform and matches the platform default). This
720 option allows you to specify a different alignment. The supplied
721 value must be greater than or equal to the default, less than or
722 equal to 4096, and must be a power of two.
723 --redzone-size=<number> [default: depends on the tool]
725 Valgrind's malloc, realloc, etc, add padding blocks before and
726 after each heap block allocated by the program being run. Such
727 padding blocks are called redzones. The default value for the
728 redzone size depends on the tool. For example, Memcheck adds and
729 protects a minimum of 16 bytes before and after each block
730 allocated by the client. This allows it to detect block underruns
731 or overruns of up to 16 bytes.
732 Increasing the redzone size makes it possible to detect overruns of
734 larger distances, but increases the amount of memory used by
735 Valgrind. Decreasing the redzone size will reduce the memory needed
736 by Valgrind but also reduces the chances of detecting
737 over/underruns, so is not recommended.
738 --xtree-memory=none|allocs|full [none]
740 Tools replacing Valgrind's malloc, realloc, etc, can optionally
741 produce an execution tree detailing which piece of code is
742 responsible for heap memory usage. See Execution Trees for a
743 detailed explanation about execution trees.
744 When set to none, no memory execution tree is produced.
746 When set to allocs, the memory execution tree gives the current
748 number of allocated bytes and the current number of allocated
749 blocks.
750 When set to full, the memory execution tree gives 6 different
752 measurements : the current number of allocated bytes and blocks
753 (same values as for allocs), the total number of allocated bytes
754 and blocks, the total number of freed bytes and blocks.
755 Note that the overhead in cpu and memory to produce an xtree
757 depends on the tool. The overhead in cpu is small for the value
758 allocs, as the information needed to produce this report is
759 maintained in any case by the tool. For massif and helgrind,
760 specifying full implies to capture a stack trace for each free
761 operation, while normally these tools only capture an allocation
762 stack trace. For Memcheck, the cpu overhead for the value full is
763 small, as this can only be used in combination with
764 --keep-stacktraces=alloc-and-free or
765 --keep-stacktraces=alloc-then-free, which already records a stack
766 trace for each free operation. The memory overhead varies between 5
767 and 10 words per unique stacktrace in the xtree, plus the memory
768 needed to record the stack trace for the free operations, if needed
769 specifically for the xtree.
770 --xtree-memory-file=<filename> [default: xtmemory.kcg.%p]
772 Specifies that Valgrind should produce the xtree memory report in
773 the specified file. Any %p or %q sequences appearing in the
774 filename are expanded in exactly the same way as they are for
775 --log-file. See the description of --log-file for details.
776 If the filename contains the extension .ms, then the produced file
778 format will be a massif output file format. If the filename
779 contains the extension .kcg or no extension is provided or
780 recognised, then the produced file format will be a callgrind
781 output format.
782 See Execution Trees for a detailed explanation about execution
784 trees formats.
785
787 These options apply to all tools, as they affect certain obscure
788 workings of the Valgrind core. Most people won't need to use them.
789 --smc-check=<none|stack|all|all-non-file> [default: all-non-file for
791 x86/amd64/s390x, stack for other archs]
792 This option controls Valgrind's detection of self-modifying code.
793 If no checking is done, when a program executes some code, then
794 overwrites it with new code, and executes the new code, Valgrind
795 will continue to execute the translations it made for the old code.
796 This will likely lead to incorrect behaviour and/or crashes.
797 For "modern" architectures -- anything that's not x86, amd64 or
799 s390x -- the default is stack. This is because a correct program
800 must take explicit action to reestablish D-I cache coherence
801 following code modification. Valgrind observes and honours such
802 actions, with the result that self-modifying code is transparently
803 handled with zero extra cost.
804 For x86, amd64 and s390x, the program is not required to notify the
806 hardware of required D-I coherence syncing. Hence the default is
807 all-non-file, which covers the normal case of generating code into
808 an anonymous (non-file-backed) mmap'd area.
809 The meanings of the four available settings are as follows. No
811 detection (none), detect self-modifying code on the stack (which is
812 used by GCC to implement nested functions) (stack), detect
813 self-modifying code everywhere (all), and detect self-modifying
814 code everywhere except in file-backed mappings (all-non-file).
815 Running with all will slow Valgrind down noticeably. Running with
817 none will rarely speed things up, since very little code gets
818 dynamically generated in most programs. The
819 VALGRIND_DISCARD_TRANSLATIONS client request is an alternative to
820 --smc-check=all and --smc-check=all-non-file that requires more
821 programmer effort but allows Valgrind to run your program faster,
822 by telling it precisely when translations need to be re-made.
823 --smc-check=all-non-file provides a cheaper but more limited
825 version of --smc-check=all. It adds checks to any translations that
826 do not originate from file-backed memory mappings. Typical
827 applications that generate code, for example JITs in web browsers,
828 generate code into anonymous mmaped areas, whereas the "fixed" code
829 of the browser always lives in file-backed mappings.
830 --smc-check=all-non-file takes advantage of this observation,
831 limiting the overhead of checking to code which is likely to be JIT
832 generated.
833 --read-inline-info=<yes|no> [default: see below]
835 When enabled, Valgrind will read information about inlined function
836 calls from DWARF3 debug info. This slows Valgrind startup and makes
837 it use more memory (typically for each inlined piece of code, 6
838 words and space for the function name), but it results in more
839 descriptive stacktraces. Currently, this functionality is enabled
840 by default only for Linux, Android and Solaris targets and only for
841 the tools Memcheck, Massif, Helgrind and DRD. Here is an example of
842 some stacktraces with --read-inline-info=no:
843 ==15380== Conditional jump or move depends on uninitialised value(s)
845 ==15380== at 0x80484EA: main (inlinfo.c:6)
846 ==15380==
847 ==15380== Conditional jump or move depends on uninitialised value(s)
848 ==15380== at 0x8048550: fun_noninline (inlinfo.c:6)
849 ==15380== by 0x804850E: main (inlinfo.c:34)
850 ==15380==
851 ==15380== Conditional jump or move depends on uninitialised value(s)
852 ==15380== at 0x8048520: main (inlinfo.c:6)
853 And here are the same errors with --read-inline-info=yes:
855 ==15377== Conditional jump or move depends on uninitialised value(s)
857 ==15377== at 0x80484EA: fun_d (inlinfo.c:6)
858 ==15377== by 0x80484EA: fun_c (inlinfo.c:14)
859 ==15377== by 0x80484EA: fun_b (inlinfo.c:20)
860 ==15377== by 0x80484EA: fun_a (inlinfo.c:26)
861 ==15377== by 0x80484EA: main (inlinfo.c:33)
862 ==15377==
863 ==15377== Conditional jump or move depends on uninitialised value(s)
864 ==15377== at 0x8048550: fun_d (inlinfo.c:6)
865 ==15377== by 0x8048550: fun_noninline (inlinfo.c:41)
866 ==15377== by 0x804850E: main (inlinfo.c:34)
867 ==15377==
868 ==15377== Conditional jump or move depends on uninitialised value(s)
869 ==15377== at 0x8048520: fun_d (inlinfo.c:6)
870 ==15377== by 0x8048520: main (inlinfo.c:35)
871 --read-var-info=<yes|no> [default: no]
873 When enabled, Valgrind will read information about variable types
874 and locations from DWARF3 debug info. This slows Valgrind startup
875 significantly and makes it use significantly more memory, but for
876 the tools that can take advantage of it (Memcheck, Helgrind, DRD)
877 it can result in more precise error messages. For example, here are
878 some standard errors issued by Memcheck:
879 ==15363== Uninitialised byte(s) found during client check request
881 ==15363== at 0x80484A9: croak (varinfo1.c:28)
882 ==15363== by 0x8048544: main (varinfo1.c:55)
883 ==15363== Address 0x80497f7 is 7 bytes inside data symbol "global_i2"
884 ==15363==
885 ==15363== Uninitialised byte(s) found during client check request
886 ==15363== at 0x80484A9: croak (varinfo1.c:28)
887 ==15363== by 0x8048550: main (varinfo1.c:56)
888 ==15363== Address 0xbea0d0cc is on thread 1's stack
889 ==15363== in frame #1, created by main (varinfo1.c:45)
890 And here are the same errors with --read-var-info=yes:
892 ==15370== Uninitialised byte(s) found during client check request
894 ==15370== at 0x80484A9: croak (varinfo1.c:28)
895 ==15370== by 0x8048544: main (varinfo1.c:55)
896 ==15370== Location 0x80497f7 is 0 bytes inside global_i2[7],
897 ==15370== a global variable declared at varinfo1.c:41
898 ==15370==
899 ==15370== Uninitialised byte(s) found during client check request
900 ==15370== at 0x80484A9: croak (varinfo1.c:28)
901 ==15370== by 0x8048550: main (varinfo1.c:56)
902 ==15370== Location 0xbeb4a0cc is 0 bytes inside local var "local"
903 ==15370== declared at varinfo1.c:46, in frame #1 of thread 1
904 --vgdb-poll=<number> [default: 5000]
906 As part of its main loop, the Valgrind scheduler will poll to check
907 if some activity (such as an external command or some input from a
908 gdb) has to be handled by gdbserver. This activity poll will be
909 done after having run the given number of basic blocks (or slightly
910 more than the given number of basic blocks). This poll is quite
911 cheap so the default value is set relatively low. You might further
912 decrease this value if vgdb cannot use ptrace system call to
913 interrupt Valgrind if all threads are (most of the time) blocked in
914 a system call.
915 --vgdb-shadow-registers=no|yes [default: no]
917 When activated, gdbserver will expose the Valgrind shadow registers
918 to GDB. With this, the value of the Valgrind shadow registers can
919 be examined or changed using GDB. Exposing shadow registers only
920 works with GDB version 7.1 or later.
921 --vgdb-prefix=<prefix> [default: /tmp/vgdb-pipe]
923 To communicate with gdb/vgdb, the Valgrind gdbserver creates 3
924 files (2 named FIFOs and a mmap shared memory file). The prefix
925 option controls the directory and prefix for the creation of these
926 files.
927 --run-libc-freeres=<yes|no> [default: yes]
929 This option is only relevant when running Valgrind on Linux.
930 The GNU C library (libc.so), which is used by all programs, may
932 allocate memory for its own uses. Usually it doesn't bother to free
933 that memory when the program ends—there would be no point, since
934 the Linux kernel reclaims all process resources when a process
935 exits anyway, so it would just slow things down.
936 The glibc authors realised that this behaviour causes leak
938 checkers, such as Valgrind, to falsely report leaks in glibc, when
939 a leak check is done at exit. In order to avoid this, they provided
940 a routine called __libc_freeres specifically to make glibc release
941 all memory it has allocated. Memcheck therefore tries to run
942 __libc_freeres at exit.
943 Unfortunately, in some very old versions of glibc, __libc_freeres
945 is sufficiently buggy to cause segmentation faults. This was
946 particularly noticeable on Red Hat 7.1. So this option is provided
947 in order to inhibit the run of __libc_freeres. If your program
948 seems to run fine on Valgrind, but segfaults at exit, you may find
949 that --run-libc-freeres=no fixes that, although at the cost of
950 possibly falsely reporting space leaks in libc.so.
951 --run-cxx-freeres=<yes|no> [default: yes]
953 This option is only relevant when running Valgrind on Linux or
954 Solaris C++ programs.
955 The GNU Standard C++ library (libstdc++.so), which is used by all
957 C++ programs compiled with g++, may allocate memory for its own
958 uses. Usually it doesn't bother to free that memory when the
959 program ends—there would be no point, since the kernel reclaims all
960 process resources when a process exits anyway, so it would just
961 slow things down.
962 The gcc authors realised that this behaviour causes leak checkers,
964 such as Valgrind, to falsely report leaks in libstdc++, when a leak
965 check is done at exit. In order to avoid this, they provided a
966 routine called __gnu_cxx::__freeres specifically to make libstdc++
967 release all memory it has allocated. Memcheck therefore tries to
968 run __gnu_cxx::__freeres at exit.
969 For the sake of flexibility and unforeseen problems with
971 __gnu_cxx::__freeres, option --run-cxx-freeres=no exists, although
972 at the cost of possibly falsely reporting space leaks in
973 libstdc++.so.
974 --sim-hints=hint1,hint2,...
976 Pass miscellaneous hints to Valgrind which slightly modify the
977 simulated behaviour in nonstandard or dangerous ways, possibly to
978 help the simulation of strange features. By default no hints are
979 enabled. Use with caution! Currently known hints are:
980 • lax-ioctls: Be very lax about ioctl handling; the only
982 assumption is that the size is correct. Doesn't require the
983 full buffer to be initialised when writing. Without this, using
984 some device drivers with a large number of strange ioctl
985 commands becomes very tiresome.
986 • fuse-compatible: Enable special handling for certain system
988 calls that may block in a FUSE file-system. This may be
989 necessary when running Valgrind on a multi-threaded program
990 that uses one thread to manage a FUSE file-system and another
991 thread to access that file-system.
992 • enable-outer: Enable some special magic needed when the program
994 being run is itself Valgrind.
995 • no-inner-prefix: Disable printing a prefix > in front of each
997 stdout or stderr output line in an inner Valgrind being run by
998 an outer Valgrind. This is useful when running Valgrind
999 regression tests in an outer/inner setup. Note that the prefix
1000 > will always be printed in front of the inner debug logging
1001 lines.
1002 • no-nptl-pthread-stackcache: This hint is only relevant when
1004 running Valgrind on Linux; it is ignored on Solaris and Mac OS
1005 X.
1006 The GNU glibc pthread library (libpthread.so), which is used by
1008 pthread programs, maintains a cache of pthread stacks. When a
1009 pthread terminates, the memory used for the pthread stack and
1010 some thread local storage related data structure are not always
1011 directly released. This memory is kept in a cache (up to a
1012 certain size), and is re-used if a new thread is started.
1013 This cache causes the helgrind tool to report some false
1015 positive race condition errors on this cached memory, as
1016 helgrind does not understand the internal glibc cache
1017 synchronisation primitives. So, when using helgrind, disabling
1018 the cache helps to avoid false positive race conditions, in
1019 particular when using thread local storage variables (e.g.
1020 variables using the __thread qualifier).
1021 When using the memcheck tool, disabling the cache ensures the
1023 memory used by glibc to handle __thread variables is directly
1024 released when a thread terminates.
1025 Note: Valgrind disables the cache using some internal knowledge
1027 of the glibc stack cache implementation and by examining the
1028 debug information of the pthread library. This technique is
1029 thus somewhat fragile and might not work for all glibc
1030 versions. This has been successfully tested with various glibc
1031 versions (e.g. 2.11, 2.16, 2.18) on various platforms.
1032 • lax-doors: (Solaris only) Be very lax about door syscall
1034 handling over unrecognised door file descriptors. Does not
1035 require that full buffer is initialised when writing. Without
1036 this, programs using libdoor(3LIB) functionality with
1037 completely proprietary semantics may report large number of
1038 false positives.
1039 • fallback-llsc: (MIPS and ARM64 only): Enables an alternative
1041 implementation of Load-Linked (LL) and Store-Conditional (SC)
1042 instructions. The standard implementation gives more correct
1043 behaviour, but can cause indefinite looping on certain
1044 processor implementations that are intolerant of extra memory
1045 references between LL and SC. So far this is known only to
1046 happen on Cavium 3 cores. You should not need to use this flag,
1047 since the relevant cores are detected at startup and the
1048 alternative implementation is automatically enabled if
1049 necessary. There is no equivalent anti-flag: you cannot
1050 force-disable the alternative implementation, if it is
1051 automatically enabled. The underlying problem exists because
1052 the "standard" implementation of LL and SC is done by copying
1053 through LL and SC instructions into the instrumented code.
1054 However, tools may insert extra instrumentation memory
1055 references in between the LL and SC instructions. These memory
1056 references are not present in the original uninstrumented code,
1057 and their presence in the instrumented code can cause the SC
1058 instructions to persistently fail, leading to indefinite
1059 looping in LL-SC blocks. The alternative implementation gives
1060 correct behaviour of LL and SC instructions between threads in
1061 a process, up to and including the ABA scenario. It also gives
1062 correct behaviour between a Valgrinded thread and a
1063 non-Valgrinded thread running in a different process, that
1064 communicate via shared memory, but only up to and including
1065 correct CAS behaviour -- in this case the ABA scenario may not
1066 be correctly handled.
1067 --fair-sched=<no|yes|try> [default: no]
1069 The --fair-sched option controls the locking mechanism used by
1070 Valgrind to serialise thread execution. The locking mechanism
1071 controls the way the threads are scheduled, and different settings
1072 give different trade-offs between fairness and performance. For
1073 more details about the Valgrind thread serialisation scheme and its
1074 impact on performance and thread scheduling, see Scheduling and
1075 Multi-Thread Performance.
1076 • The value --fair-sched=yes activates a fair scheduler. In
1078 short, if multiple threads are ready to run, the threads will
1079 be scheduled in a round robin fashion. This mechanism is not
1080 available on all platforms or Linux versions. If not available,
1081 using --fair-sched=yes will cause Valgrind to terminate with an
1082 error.
1083 You may find this setting improves overall responsiveness if
1085 you are running an interactive multithreaded program, for
1086 example a web browser, on Valgrind.
1087 • The value --fair-sched=try activates fair scheduling if
1089 available on the platform. Otherwise, it will automatically
1090 fall back to --fair-sched=no.
1091 • The value --fair-sched=no activates a scheduler which does not
1093 guarantee fairness between threads ready to run, but which in
1094 general gives the highest performance.
1095 --kernel-variant=variant1,variant2,...
1097 Handle system calls and ioctls arising from minor variants of the
1098 default kernel for this platform. This is useful for running on
1099 hacked kernels or with kernel modules which support nonstandard
1100 ioctls, for example. Use with caution. If you don't understand what
1101 this option does then you almost certainly don't need it. Currently
1102 known variants are:
1103 • bproc: support the sys_broc system call on x86. This is for
1105 running on BProc, which is a minor variant of standard Linux
1106 which is sometimes used for building clusters.
1107 • android-no-hw-tls: some versions of the Android emulator for
1109 ARM do not provide a hardware TLS (thread-local state)
1110 register, and Valgrind crashes at startup. Use this variant to
1111 select software support for TLS.
1112 • android-gpu-sgx5xx: use this to support handling of proprietary
1114 ioctls for the PowerVR SGX 5XX series of GPUs on Android
1115 devices. Failure to select this does not cause stability
1116 problems, but may cause Memcheck to report false errors after
1117 the program performs GPU-specific ioctls.
1118 • android-gpu-adreno3xx: similarly, use this to support handling
1120 of proprietary ioctls for the Qualcomm Adreno 3XX series of
1121 GPUs on Android devices.
1122 --merge-recursive-frames=<number> [default: 0]
1124 Some recursive algorithms, for example balanced binary tree
1125 implementations, create many different stack traces, each
1126 containing cycles of calls. A cycle is defined as two identical
1127 program counter values separated by zero or more other program
1128 counter values. Valgrind may then use a lot of memory to store all
1129 these stack traces. This is a poor use of memory considering that
1130 such stack traces contain repeated uninteresting recursive calls
1131 instead of more interesting information such as the function that
1132 has initiated the recursive call.
1133 The option --merge-recursive-frames=<number> instructs Valgrind to
1135 detect and merge recursive call cycles having a size of up to
1136 <number> frames. When such a cycle is detected, Valgrind records
1137 the cycle in the stack trace as a unique program counter.
1138 The value 0 (the default) causes no recursive call merging. A value
1140 of 1 will cause stack traces of simple recursive algorithms (for
1141 example, a factorial implementation) to be collapsed. A value of 2
1142 will usually be needed to collapse stack traces produced by
1143 recursive algorithms such as binary trees, quick sort, etc. Higher
1144 values might be needed for more complex recursive algorithms.
1145 Note: recursive calls are detected by analysis of program counter
1147 values. They are not detected by looking at function names.
1148 --num-transtab-sectors=<number> [default: 6 for Android platforms, 16
1150 for all others]
1151 Valgrind translates and instruments your program's machine code in
1152 small fragments (basic blocks). The translations are stored in a
1153 translation cache that is divided into a number of sections
1154 (sectors). If the cache is full, the sector containing the oldest
1155 translations is emptied and reused. If these old translations are
1156 needed again, Valgrind must re-translate and re-instrument the
1157 corresponding machine code, which is expensive. If the "executed
1158 instructions" working set of a program is big, increasing the
1159 number of sectors may improve performance by reducing the number of
1160 re-translations needed. Sectors are allocated on demand. Once
1161 allocated, a sector can never be freed, and occupies considerable
1162 space, depending on the tool and the value of
1163 --avg-transtab-entry-size (about 40 MB per sector for Memcheck).
1164 Use the option --stats=yes to obtain precise information about the
1165 memory used by a sector and the allocation and recycling of
1166 sectors.
1167 --avg-transtab-entry-size=<number> [default: 0, meaning use tool
1169 provided default]
1170 Average size of translated basic block. This average size is used
1171 to dimension the size of a sector. Each tool provides a default
1172 value to be used. If this default value is too small, the
1173 translation sectors will become full too quickly. If this default
1174 value is too big, a significant part of the translation sector
1175 memory will be unused. Note that the average size of a basic block
1176 translation depends on the tool, and might depend on tool options.
1177 For example, the memcheck option --track-origins=yes increases the
1178 size of the basic block translations. Use --avg-transtab-entry-size
1179 to tune the size of the sectors, either to gain memory or to avoid
1180 too many retranslations.
1181 --aspace-minaddr=<address> [default: depends on the platform]
1183 To avoid potential conflicts with some system libraries, Valgrind
1184 does not use the address space below --aspace-minaddr value,
1185 keeping it reserved in case a library specifically requests memory
1186 in this region. So, some "pessimistic" value is guessed by Valgrind
1187 depending on the platform. On linux, by default, Valgrind avoids
1188 using the first 64MB even if typically there is no conflict in this
1189 complete zone. You can use the option --aspace-minaddr to have your
1190 memory hungry application benefitting from more of this lower
1191 memory. On the other hand, if you encounter a conflict, increasing
1192 aspace-minaddr value might solve it. Conflicts will typically
1193 manifest themselves with mmap failures in the low range of the
1194 address space. The provided address must be page aligned and must
1195 be equal or bigger to 0x1000 (4KB). To find the default value on
1196 your platform, do something such as valgrind -d -d date 2>&1 | grep
1197 -i minaddr. Values lower than 0x10000 (64KB) are known to create
1198 problems on some distributions.
1199 --valgrind-stacksize=<number> [default: 1MB]
1201 For each thread, Valgrind needs its own 'private' stack. The
1202 default size for these stacks is largely dimensioned, and so should
1203 be sufficient in most cases. In case the size is too small,
1204 Valgrind will segfault. Before segfaulting, a warning might be
1205 produced by Valgrind when approaching the limit.
1206 Use the option --valgrind-stacksize if such an (unlikely) warning
1208 is produced, or Valgrind dies due to a segmentation violation. Such
1209 segmentation violations have been seen when demangling huge C++
1210 symbols.
1211 If your application uses many threads and needs a lot of memory,
1213 you can gain some memory by reducing the size of these Valgrind
1214 stacks using the option --valgrind-stacksize.
1215 --show-emwarns=<yes|no> [default: no]
1217 When enabled, Valgrind will emit warnings about its CPU emulation
1218 in certain cases. These are usually not interesting.
1219 --require-text-symbol=:sonamepatt:fnnamepatt
1221 When a shared object whose soname matches sonamepatt is loaded into
1222 the process, examine all the text symbols it exports. If none of
1223 those match fnnamepatt, print an error message and abandon the run.
1224 This makes it possible to ensure that the run does not continue
1225 unless a given shared object contains a particular function name.
1226 Both sonamepatt and fnnamepatt can be written using the usual ?
1228 and * wildcards. For example: ":*libc.so*:foo?bar". You may use
1229 characters other than a colon to separate the two patterns. It is
1230 only important that the first character and the separator character
1231 are the same. For example, the above example could also be written
1232 "Q*libc.so*Qfoo?bar". Multiple
1233 --require-text-symbol flags are allowed, in which case shared
1234 objects that are loaded into the process will be checked against
1235 all of them.
1236 The purpose of this is to support reliable usage of marked-up
1238 libraries. For example, suppose we have a version of GCC's
1239 libgomp.so which has been marked up with annotations to support
1240 Helgrind. It is only too easy and confusing to load the wrong,
1241 un-annotated libgomp.so into the application. So the idea is: add a
1242 text symbol in the marked-up library, for example
1243 annotated_for_helgrind_3_6, and then give the flag
1244 --require-text-symbol=:*libgomp*so*:annotated_for_helgrind_3_6 so
1245 that when libgomp.so is loaded, Valgrind scans its symbol table,
1246 and if the symbol isn't present the run is aborted, rather than
1247 continuing silently with the un-marked-up library. Note that you
1248 should put the entire flag in quotes to stop shells expanding up
1249 the * and ? wildcards.
1250 --soname-synonyms=syn1=pattern1,syn2=pattern2,...
1252 When a shared library is loaded, Valgrind checks for functions in
1253 the library that must be replaced or wrapped. For example, Memcheck
1254 replaces some string and memory functions (strchr, strlen, strcpy,
1255 memchr, memcpy, memmove, etc.) with its own versions. Such
1256 replacements are normally done only in shared libraries whose
1257 soname matches a predefined soname pattern (e.g. libc.so* on
1258 linux). By default, no replacement is done for a statically linked
1259 binary or for alternative libraries, except for the allocation
1260 functions (malloc, free, calloc, memalign, realloc, operator new,
1261 operator delete, etc.) Such allocation functions are intercepted by
1262 default in any shared library or in the executable if they are
1263 exported as global symbols. This means that if a replacement
1264 allocation library such as tcmalloc is found, its functions are
1265 also intercepted by default. In some cases, the replacements allow
1266 --soname-synonyms to specify one additional synonym pattern, giving
1267 flexibility in the replacement. Or to prevent interception of all
1268 public allocation symbols.
1269 Currently, this flexibility is only allowed for the malloc related
1271 functions, using the synonym somalloc. This synonym is usable for
1272 all tools doing standard replacement of malloc related functions
1273 (e.g. memcheck, helgrind, drd, massif, dhat).
1274 • Alternate malloc library: to replace the malloc related
1276 functions in a specific alternate library with soname
1277 mymalloclib.so (and not in any others), give the option
1278 --soname-synonyms=somalloc=mymalloclib.so. A pattern can be
1279 used to match multiple libraries sonames. For example,
1280 --soname-synonyms=somalloc=*tcmalloc* will match the soname of
1281 all variants of the tcmalloc library (native, debug, profiled,
1282 ... tcmalloc variants).
1283 Note: the soname of a elf shared library can be retrieved using
1285 the readelf utility.
1286 • Replacements in a statically linked library are done by using
1288 the NONE pattern. For example, if you link with libtcmalloc.a,
1289 and only want to intercept the malloc related functions in the
1290 executable (and standard libraries) themselves, but not any
1291 other shared libraries, you can give the option
1292 --soname-synonyms=somalloc=NONE. Note that a NONE pattern will
1293 match the main executable and any shared library having no
1294 soname.
1295 • To run a "default" Firefox build for Linux, in which JEMalloc
1297 is linked in to the main executable, use
1298 --soname-synonyms=somalloc=NONE.
1299 • To only intercept allocation symbols in the default system
1301 libraries, but not in any other shared library or the
1302 executable defining public malloc or operator new related
1303 functions use a non-existing library name like
1304 --soname-synonyms=somalloc=nouserintercepts (where
1305 nouserintercepts can be any non-existing library name).
1306 • Shared library of the dynamic (runtime) linker is excluded from
1308 searching for global public symbols, such as those for the
1309 malloc related functions (identified by somalloc synonym).
1310 --progress-interval=<number> [default: 0, meaning 'disabled']
1312 This is an enhancement to Valgrind's debugging output. It is
1313 unlikely to be of interest to end users.
1314 When number is set to a non-zero value, Valgrind will print a
1316 one-line progress summary every number seconds. Valid settings for
1317 number are between 0 and 3600 inclusive. Here's some example output
1318 with number set to 10:
1319 PROGRESS: U 110s, W 113s, 97.3% CPU, EvC 414.79M, TIn 616.7k, TOut 0.5k, #thr 67
1321 PROGRESS: U 120s, W 124s, 96.8% CPU, EvC 505.27M, TIn 636.6k, TOut 3.0k, #thr 64
1322 PROGRESS: U 130s, W 134s, 97.0% CPU, EvC 574.90M, TIn 657.5k, TOut 3.0k, #thr 63
1323 Each line shows:
1325 • U: total user time
1327 • W: total wallclock time
1329 • CPU: overall average cpu use
1331 • EvC: number of event checks. An event check is a backwards
1333 branch in the simulated program, so this is a measure of
1334 forward progress of the program
1335 • TIn: number of code blocks instrumented by the JIT
1337 • TOut: number of instrumented code blocks that have been thrown
1339 away
1340 • #thr: number of threads in the program
1342 From the progress of these, it is possible to observe:
1344 • when the program is compute bound (TIn rises slowly, EvC rises
1346 rapidly)
1347 • when the program is in a spinloop (TIn/TOut fixed, EvC rises
1349 rapidly)
1350 • when the program is JIT-bound (TIn rises rapidly)
1352 • when the program is rapidly discarding code (TOut rises
1354 rapidly)
1355 • when the program is about to achieve some expected state (EvC
1357 arrives at some value you expect)
1358 • when the program is idling (U rises more slowly than W)
1360
1363 There are also some options for debugging Valgrind itself. You
1364 shouldn't need to use them in the normal run of things. If you wish to
1365 see the list, use the --help-debug option.
1366
1368 --leak-check=<no|summary|yes|full> [default: summary]
1369 When enabled, search for memory leaks when the client program
1370 finishes. If set to summary, it says how many leaks occurred. If
1371 set to full or yes, each individual leak will be shown in detail
1372 and/or counted as an error, as specified by the options
1373 --show-leak-kinds and --errors-for-leak-kinds.
1374 If --xml=yes is given, memcheck will automatically use the value
1376 --leak-check=full. You can use --show-leak-kinds=none to reduce the
1377 size of the xml output if you are not interested in the leak
1378 results.
1379 --leak-resolution=<low|med|high> [default: high]
1381 When doing leak checking, determines how willing Memcheck is to
1382 consider different backtraces to be the same for the purposes of
1383 merging multiple leaks into a single leak report. When set to low,
1384 only the first two entries need match. When med, four entries have
1385 to match. When high, all entries need to match.
1386 For hardcore leak debugging, you probably want to use
1388 --leak-resolution=high together with --num-callers=40 or some such
1389 large number.
1390 Note that the --leak-resolution setting does not affect Memcheck's
1392 ability to find leaks. It only changes how the results are
1393 presented.
1394 --show-leak-kinds=<set> [default: definite,possible]
1396 Specifies the leak kinds to show in a full leak search, in one of
1397 the following ways:
1398 • a comma separated list of one or more of definite indirect
1400 possible reachable.
1401 • all to specify the complete set (all leak kinds). It is
1403 equivalent to
1404 --show-leak-kinds=definite,indirect,possible,reachable.
1405 • none for the empty set.
1407 --errors-for-leak-kinds=<set> [default: definite,possible]
1409 Specifies the leak kinds to count as errors in a full leak search.
1410 The <set> is specified similarly to --show-leak-kinds
1411 --leak-check-heuristics=<set> [default: all]
1413 Specifies the set of leak check heuristics to be used during leak
1414 searches. The heuristics control which interior pointers to a block
1415 cause it to be considered as reachable. The heuristic set is
1416 specified in one of the following ways:
1417 • a comma separated list of one or more of stdstring length64
1419 newarray multipleinheritance.
1420 • all to activate the complete set of heuristics. It is
1422 equivalent to
1423 --leak-check-heuristics=stdstring,length64,newarray,multipleinheritance.
1424 • none for the empty set.
1426 Note that these heuristics are dependent on the layout of the
1428 objects produced by the C++ compiler. They have been tested with
1429 some gcc versions (e.g. 4.4 and 4.7). They might not work properly
1430 with other C++ compilers.
1431 --show-reachable=<yes|no> , --show-possibly-lost=<yes|no>
1433 These options provide an alternative way to specify the leak kinds
1434 to show:
1435 • --show-reachable=no --show-possibly-lost=yes is equivalent to
1437 --show-leak-kinds=definite,possible.
1438 • --show-reachable=no --show-possibly-lost=no is equivalent to
1440 --show-leak-kinds=definite.
1441 • --show-reachable=yes is equivalent to --show-leak-kinds=all.
1443 Note that --show-possibly-lost=no has no effect if
1445 --show-reachable=yes is specified.
1446 --xtree-leak=<no|yes> [no]
1448 If set to yes, the results for the leak search done at exit will be
1449 output in a 'Callgrind Format' execution tree file. Note that this
1450 automatically sets the options --leak-check=full and
1451 --show-leak-kinds=all, to allow xtree visualisation tools such as
1452 kcachegrind to select what kind to leak to visualise. The produced
1453 file will contain the following events:
1454 • RB : Reachable Bytes
1456 • PB : Possibly lost Bytes
1458 • IB : Indirectly lost Bytes
1460 • DB : Definitely lost Bytes (direct plus indirect)
1462 • DIB : Definitely Indirectly lost Bytes (subset of DB)
1464 • RBk : reachable Blocks
1466 • PBk : Possibly lost Blocks
1468 • IBk : Indirectly lost Blocks
1470 • DBk : Definitely lost Blocks
1472 The increase or decrease for all events above will also be output
1474 in the file to provide the delta (increase or decrease) between 2
1475 successive leak searches. For example, iRB is the increase of the
1476 RB event, dPBk is the decrease of PBk event. The values for the
1477 increase and decrease events will be zero for the first leak search
1478 done.
1479 See Execution Trees for a detailed explanation about execution
1481 trees.
1482 --xtree-leak-file=<filename> [default: xtleak.kcg.%p]
1484 Specifies that Valgrind should produce the xtree leak report in the
1485 specified file. Any %p, %q or %n sequences appearing in the
1486 filename are expanded in exactly the same way as they are for
1487 --log-file. See the description of --log-file for details.
1488 See Execution Trees for a detailed explanation about execution
1490 trees formats.
1491 --undef-value-errors=<yes|no> [default: yes]
1493 Controls whether Memcheck reports uses of undefined value errors.
1494 Set this to no if you don't want to see undefined value errors. It
1495 also has the side effect of speeding up Memcheck somewhat.
1496 AddrCheck (removed in Valgrind 3.1.0) functioned like Memcheck with
1497 --undef-value-errors=no.
1498 --track-origins=<yes|no> [default: no]
1500 Controls whether Memcheck tracks the origin of uninitialised
1501 values. By default, it does not, which means that although it can
1502 tell you that an uninitialised value is being used in a dangerous
1503 way, it cannot tell you where the uninitialised value came from.
1504 This often makes it difficult to track down the root problem.
1505 When set to yes, Memcheck keeps track of the origins of all
1507 uninitialised values. Then, when an uninitialised value error is
1508 reported, Memcheck will try to show the origin of the value. An
1509 origin can be one of the following four places: a heap block, a
1510 stack allocation, a client request, or miscellaneous other sources
1511 (eg, a call to brk).
1512 For uninitialised values originating from a heap block, Memcheck
1514 shows where the block was allocated. For uninitialised values
1515 originating from a stack allocation, Memcheck can tell you which
1516 function allocated the value, but no more than that -- typically it
1517 shows you the source location of the opening brace of the function.
1518 So you should carefully check that all of the function's local
1519 variables are initialised properly.
1520 Performance overhead: origin tracking is expensive. It halves
1522 Memcheck's speed and increases memory use by a minimum of 100MB,
1523 and possibly more. Nevertheless it can drastically reduce the
1524 effort required to identify the root cause of uninitialised value
1525 errors, and so is often a programmer productivity win, despite
1526 running more slowly.
1527 Accuracy: Memcheck tracks origins quite accurately. To avoid very
1529 large space and time overheads, some approximations are made. It is
1530 possible, although unlikely, that Memcheck will report an incorrect
1531 origin, or not be able to identify any origin.
1532 Note that the combination --track-origins=yes and
1534 --undef-value-errors=no is nonsensical. Memcheck checks for and
1535 rejects this combination at startup.
1536 --partial-loads-ok=<yes|no> [default: yes]
1538 Controls how Memcheck handles 32-, 64-, 128- and 256-bit naturally
1539 aligned loads from addresses for which some bytes are addressable
1540 and others are not. When yes, such loads do not produce an address
1541 error. Instead, loaded bytes originating from illegal addresses are
1542 marked as uninitialised, and those corresponding to legal addresses
1543 are handled in the normal way.
1544 When no, loads from partially invalid addresses are treated the
1546 same as loads from completely invalid addresses: an illegal-address
1547 error is issued, and the resulting bytes are marked as initialised.
1548 Note that code that behaves in this way is in violation of the ISO
1550 C/C++ standards, and should be considered broken. If at all
1551 possible, such code should be fixed.
1552 --expensive-definedness-checks=<no|auto|yes> [default: auto]
1554 Controls whether Memcheck should employ more precise but also more
1555 expensive (time consuming) instrumentation when checking the
1556 definedness of certain values. In particular, this affects the
1557 instrumentation of integer adds, subtracts and equality
1558 comparisons.
1559 Selecting --expensive-definedness-checks=yes causes Memcheck to use
1561 the most accurate analysis possible. This minimises false error
1562 rates but can cause up to 30% performance degradation.
1563 Selecting --expensive-definedness-checks=no causes Memcheck to use
1565 the cheapest instrumentation possible. This maximises performance
1566 but will normally give an unusably high false error rate.
1567 The default setting, --expensive-definedness-checks=auto, is
1569 strongly recommended. This causes Memcheck to use the minimum of
1570 expensive instrumentation needed to achieve the same false error
1571 rate as --expensive-definedness-checks=yes. It also enables an
1572 instrumentation-time analysis pass which aims to further reduce the
1573 costs of accurate instrumentation. Overall, the performance loss is
1574 generally around 5% relative to --expensive-definedness-checks=no,
1575 although this is strongly workload dependent. Note that the exact
1576 instrumentation settings in this mode are architecture dependent.
1577 --keep-stacktraces=alloc|free|alloc-and-free|alloc-then-free|none
1579 [default: alloc-and-free]
1580 Controls which stack trace(s) to keep for malloc'd and/or free'd
1581 blocks.
1582 With alloc-then-free, a stack trace is recorded at allocation time,
1584 and is associated with the block. When the block is freed, a second
1585 stack trace is recorded, and this replaces the allocation stack
1586 trace. As a result, any "use after free" errors relating to this
1587 block can only show a stack trace for where the block was freed.
1588 With alloc-and-free, both allocation and the deallocation stack
1590 traces for the block are stored. Hence a "use after free" error
1591 will show both, which may make the error easier to diagnose.
1592 Compared to alloc-then-free, this setting slightly increases
1593 Valgrind's memory use as the block contains two references instead
1594 of one.
1595 With alloc, only the allocation stack trace is recorded (and
1597 reported). With free, only the deallocation stack trace is recorded
1598 (and reported). These values somewhat decrease Valgrind's memory
1599 and cpu usage. They can be useful depending on the error types you
1600 are searching for and the level of detail you need to analyse them.
1601 For example, if you are only interested in memory leak errors, it
1602 is sufficient to record the allocation stack traces.
1603 With none, no stack traces are recorded for malloc and free
1605 operations. If your program allocates a lot of blocks and/or
1606 allocates/frees from many different stack traces, this can
1607 significantly decrease cpu and/or memory required. Of course, few
1608 details will be reported for errors related to heap blocks.
1609 Note that once a stack trace is recorded, Valgrind keeps the stack
1611 trace in memory even if it is not referenced by any block. Some
1612 programs (for example, recursive algorithms) can generate a huge
1613 number of stack traces. If Valgrind uses too much memory in such
1614 circumstances, you can reduce the memory required with the options
1615 --keep-stacktraces and/or by using a smaller value for the option
1616 --num-callers.
1617 If you want to use --xtree-memory=full memory profiling (see
1619 Execution Trees), then you cannot specify --keep-stacktraces=free
1620 or --keep-stacktraces=none.
1621 --freelist-vol=<number> [default: 20000000]
1623 When the client program releases memory using free (in C) or delete
1624 (C++), that memory is not immediately made available for
1625 re-allocation. Instead, it is marked inaccessible and placed in a
1626 queue of freed blocks. The purpose is to defer as long as possible
1627 the point at which freed-up memory comes back into circulation.
1628 This increases the chance that Memcheck will be able to detect
1629 invalid accesses to blocks for some significant period of time
1630 after they have been freed.
1631 This option specifies the maximum total size, in bytes, of the
1633 blocks in the queue. The default value is twenty million bytes.
1634 Increasing this increases the total amount of memory used by
1635 Memcheck but may detect invalid uses of freed blocks which would
1636 otherwise go undetected.
1637 --freelist-big-blocks=<number> [default: 1000000]
1639 When making blocks from the queue of freed blocks available for
1640 re-allocation, Memcheck will in priority re-circulate the blocks
1641 with a size greater or equal to --freelist-big-blocks. This ensures
1642 that freeing big blocks (in particular freeing blocks bigger than
1643 --freelist-vol) does not immediately lead to a re-circulation of
1644 all (or a lot of) the small blocks in the free list. In other
1645 words, this option increases the likelihood to discover dangling
1646 pointers for the "small" blocks, even when big blocks are freed.
1647 Setting a value of 0 means that all the blocks are re-circulated in
1649 a FIFO order.
1650 --workaround-gcc296-bugs=<yes|no> [default: no]
1652 When enabled, assume that reads and writes some small distance
1653 below the stack pointer are due to bugs in GCC 2.96, and does not
1654 report them. The "small distance" is 256 bytes by default. Note
1655 that GCC 2.96 is the default compiler on some ancient Linux
1656 distributions (RedHat 7.X) and so you may need to use this option.
1657 Do not use it if you do not have to, as it can cause real errors to
1658 be overlooked. A better alternative is to use a more recent GCC in
1659 which this bug is fixed.
1660 You may also need to use this option when working with GCC 3.X or
1662 4.X on 32-bit PowerPC Linux. This is because GCC generates code
1663 which occasionally accesses below the stack pointer, particularly
1664 for floating-point to/from integer conversions. This is in
1665 violation of the 32-bit PowerPC ELF specification, which makes no
1666 provision for locations below the stack pointer to be accessible.
1667 This option is deprecated as of version 3.12 and may be removed
1669 from future versions. You should instead use
1670 --ignore-range-below-sp to specify the exact range of offsets below
1671 the stack pointer that should be ignored. A suitable equivalent is
1672 --ignore-range-below-sp=1024-1.
1673 --ignore-range-below-sp=<number>-<number>
1675 This is a more general replacement for the deprecated
1676 --workaround-gcc296-bugs option. When specified, it causes Memcheck
1677 not to report errors for accesses at the specified offsets below
1678 the stack pointer. The two offsets must be positive decimal numbers
1679 and -- somewhat counterintuitively -- the first one must be larger,
1680 in order to imply a non-wraparound address range to ignore. For
1681 example, to ignore 4 byte accesses at 8192 bytes below the stack
1682 pointer, use --ignore-range-below-sp=8192-8189. Only one range may
1683 be specified.
1684 --show-mismatched-frees=<yes|no> [default: yes]
1686 When enabled, Memcheck checks that heap blocks are deallocated
1687 using a function that matches the allocating function. That is, it
1688 expects free to be used to deallocate blocks allocated by malloc,
1689 delete for blocks allocated by new, and delete[] for blocks
1690 allocated by new[]. If a mismatch is detected, an error is
1691 reported. This is in general important because in some
1692 environments, freeing with a non-matching function can cause
1693 crashes.
1694 There is however a scenario where such mismatches cannot be
1696 avoided. That is when the user provides implementations of
1697 new/new[] that call malloc and of delete/delete[] that call free,
1698 and these functions are asymmetrically inlined. For example,
1699 imagine that delete[] is inlined but new[] is not. The result is
1700 that Memcheck "sees" all delete[] calls as direct calls to free,
1701 even when the program source contains no mismatched calls.
1702 This causes a lot of confusing and irrelevant error reports.
1704 --show-mismatched-frees=no disables these checks. It is not
1705 generally advisable to disable them, though, because you may miss
1706 real errors as a result.
1707 --ignore-ranges=0xPP-0xQQ[,0xRR-0xSS]
1709 Any ranges listed in this option (and multiple ranges can be
1710 specified, separated by commas) will be ignored by Memcheck's
1711 addressability checking.
1712 --malloc-fill=<hexnumber>
1714 Fills blocks allocated by malloc, new, etc, but not by calloc, with
1715 the specified byte. This can be useful when trying to shake out
1716 obscure memory corruption problems. The allocated area is still
1717 regarded by Memcheck as undefined -- this option only affects its
1718 contents. Note that --malloc-fill does not affect a block of memory
1719 when it is used as argument to client requests
1720 VALGRIND_MEMPOOL_ALLOC or VALGRIND_MALLOCLIKE_BLOCK.
1721 --free-fill=<hexnumber>
1723 Fills blocks freed by free, delete, etc, with the specified byte
1724 value. This can be useful when trying to shake out obscure memory
1725 corruption problems. The freed area is still regarded by Memcheck
1726 as not valid for access -- this option only affects its contents.
1727 Note that --free-fill does not affect a block of memory when it is
1728 used as argument to client requests VALGRIND_MEMPOOL_FREE or
1729 VALGRIND_FREELIKE_BLOCK.
1730
1732 --I1=<size>,<associativity>,<line size>
1733 Specify the size, associativity and line size of the level 1
1734 instruction cache.
1735 --D1=<size>,<associativity>,<line size>
1737 Specify the size, associativity and line size of the level 1 data
1738 cache.
1739 --LL=<size>,<associativity>,<line size>
1741 Specify the size, associativity and line size of the last-level
1742 cache.
1743 --cache-sim=no|yes [yes]
1745 Enables or disables collection of cache access and miss counts.
1746 --branch-sim=no|yes [no]
1748 Enables or disables collection of branch instruction and
1749 misprediction counts. By default this is disabled as it slows
1750 Cachegrind down by approximately 25%. Note that you cannot specify
1751 --cache-sim=no and --branch-sim=no together, as that would leave
1752 Cachegrind with no information to collect.
1753 --cachegrind-out-file=<file>
1755 Write the profile data to file rather than to the default output
1756 file, cachegrind.out.<pid>. The %p and %q format specifiers can be
1757 used to embed the process ID and/or the contents of an environment
1758 variable in the name, as is the case for the core option
1759 --log-file.
1760
1762 --callgrind-out-file=<file>
1763 Write the profile data to file rather than to the default output
1764 file, callgrind.out.<pid>. The %p and %q format specifiers can be
1765 used to embed the process ID and/or the contents of an environment
1766 variable in the name, as is the case for the core option
1767 --log-file. When multiple dumps are made, the file name is modified
1768 further; see below.
1769 --dump-line=<no|yes> [default: yes]
1771 This specifies that event counting should be performed at source
1772 line granularity. This allows source annotation for sources which
1773 are compiled with debug information (-g).
1774 --dump-instr=<no|yes> [default: no]
1776 This specifies that event counting should be performed at
1777 per-instruction granularity. This allows for assembly code
1778 annotation. Currently the results can only be displayed by
1779 KCachegrind.
1780 --compress-strings=<no|yes> [default: yes]
1782 This option influences the output format of the profile data. It
1783 specifies whether strings (file and function names) should be
1784 identified by numbers. This shrinks the file, but makes it more
1785 difficult for humans to read (which is not recommended in any
1786 case).
1787 --compress-pos=<no|yes> [default: yes]
1789 This option influences the output format of the profile data. It
1790 specifies whether numerical positions are always specified as
1791 absolute values or are allowed to be relative to previous numbers.
1792 This shrinks the file size.
1793 --combine-dumps=<no|yes> [default: no]
1795 When enabled, when multiple profile data parts are to be generated
1796 these parts are appended to the same output file. Not recommended.
1797 --dump-every-bb=<count> [default: 0, never]
1799 Dump profile data every count basic blocks. Whether a dump is
1800 needed is only checked when Valgrind's internal scheduler is run.
1801 Therefore, the minimum setting useful is about 100000. The count is
1802 a 64-bit value to make long dump periods possible.
1803 --dump-before=<function>
1805 Dump when entering function.
1806 --zero-before=<function>
1808 Zero all costs when entering function.
1809 --dump-after=<function>
1811 Dump when leaving function.
1812 --instr-atstart=<yes|no> [default: yes]
1814 Specify if you want Callgrind to start simulation and profiling
1815 from the beginning of the program. When set to no, Callgrind will
1816 not be able to collect any information, including calls, but it
1817 will have at most a slowdown of around 4, which is the minimum
1818 Valgrind overhead. Instrumentation can be interactively enabled via
1819 callgrind_control -i on.
1820 Note that the resulting call graph will most probably not contain
1822 main, but will contain all the functions executed after
1823 instrumentation was enabled. Instrumentation can also be
1824 programmatically enabled/disabled. See the Callgrind include file
1825 callgrind.h for the macro you have to use in your source code.
1826 For cache simulation, results will be less accurate when switching
1828 on instrumentation later in the program run, as the simulator
1829 starts with an empty cache at that moment. Switch on event
1830 collection later to cope with this error.
1831 --collect-atstart=<yes|no> [default: yes]
1833 Specify whether event collection is enabled at beginning of the
1834 profile run.
1835 To only look at parts of your program, you have two possibilities:
1837 1. Zero event counters before entering the program part you want
1839 to profile, and dump the event counters to a file after leaving
1840 that program part.
1841 2. Switch on/off collection state as needed to only see event
1843 counters happening while inside of the program part you want to
1844 profile.
1845 The second option can be used if the program part you want to
1847 profile is called many times. Option 1, i.e. creating a lot of
1848 dumps is not practical here.
1849 Collection state can be toggled at entry and exit of a given
1851 function with the option --toggle-collect. If you use this option,
1852 collection state should be disabled at the beginning. Note that the
1853 specification of --toggle-collect implicitly sets
1854 --collect-state=no.
1855 Collection state can be toggled also by inserting the client
1857 request CALLGRIND_TOGGLE_COLLECT ; at the needed code positions.
1858 --toggle-collect=<function>
1860 Toggle collection on entry/exit of function.
1861 --collect-jumps=<no|yes> [default: no]
1863 This specifies whether information for (conditional) jumps should
1864 be collected. As above, callgrind_annotate currently is not able to
1865 show you the data. You have to use KCachegrind to get jump arrows
1866 in the annotated code.
1867 --collect-systime=<no|yes|msec|usec|nsec> [default: no]
1869 This specifies whether information for system call times should be
1870 collected.
1871 The value no indicates to record no system call information.
1873 The other values indicate to record the number of system calls done
1875 (sysCount event) and the elapsed time (sysTime event) spent in
1876 system calls. The --collect-systime value gives the unit used for
1877 sysTime : milli seconds, micro seconds or nano seconds. With the
1878 value nsec, callgrind also records the cpu time spent during system
1879 calls (sysCpuTime).
1880 The value yes is a synonym of msec. The value nsec is not supported
1882 on Darwin.
1883 --collect-bus=<no|yes> [default: no]
1885 This specifies whether the number of global bus events executed
1886 should be collected. The event type "Ge" is used for these events.
1887 --cache-sim=<yes|no> [default: no]
1889 Specify if you want to do full cache simulation. By default, only
1890 instruction read accesses will be counted ("Ir"). With cache
1891 simulation, further event counters are enabled: Cache misses on
1892 instruction reads ("I1mr"/"ILmr"), data read accesses ("Dr") and
1893 related cache misses ("D1mr"/"DLmr"), data write accesses ("Dw")
1894 and related cache misses ("D1mw"/"DLmw"). For more information, see
1895 Cachegrind: a cache and branch-prediction profiler.
1896 --branch-sim=<yes|no> [default: no]
1898 Specify if you want to do branch prediction simulation. Further
1899 event counters are enabled: Number of executed conditional branches
1900 and related predictor misses ("Bc"/"Bcm"), executed indirect jumps
1901 and related misses of the jump address predictor ("Bi"/"Bim").
1902
1904 --free-is-write=no|yes [default: no]
1905 When enabled (not the default), Helgrind treats freeing of heap
1906 memory as if the memory was written immediately before the free.
1907 This exposes races where memory is referenced by one thread, and
1908 freed by another, but there is no observable synchronisation event
1909 to ensure that the reference happens before the free.
1910 This functionality is new in Valgrind 3.7.0, and is regarded as
1912 experimental. It is not enabled by default because its interaction
1913 with custom memory allocators is not well understood at present.
1914 User feedback is welcomed.
1915 --track-lockorders=no|yes [default: yes]
1917 When enabled (the default), Helgrind performs lock order
1918 consistency checking. For some buggy programs, the large number of
1919 lock order errors reported can become annoying, particularly if
1920 you're only interested in race errors. You may therefore find it
1921 helpful to disable lock order checking.
1922 --history-level=none|approx|full [default: full]
1924 --history-level=full (the default) causes Helgrind collects enough
1925 information about "old" accesses that it can produce two stack
1926 traces in a race report -- both the stack trace for the current
1927 access, and the trace for the older, conflicting access. To limit
1928 memory usage, "old" accesses stack traces are limited to a maximum
1929 of 8 entries, even if --num-callers value is bigger.
1930 Collecting such information is expensive in both speed and memory,
1932 particularly for programs that do many inter-thread synchronisation
1933 events (locks, unlocks, etc). Without such information, it is more
1934 difficult to track down the root causes of races. Nonetheless, you
1935 may not need it in situations where you just want to check for the
1936 presence or absence of races, for example, when doing regression
1937 testing of a previously race-free program.
1938 --history-level=none is the opposite extreme. It causes Helgrind
1940 not to collect any information about previous accesses. This can be
1941 dramatically faster than --history-level=full.
1942 --history-level=approx provides a compromise between these two
1944 extremes. It causes Helgrind to show a full trace for the later
1945 access, and approximate information regarding the earlier access.
1946 This approximate information consists of two stacks, and the
1947 earlier access is guaranteed to have occurred somewhere between
1948 program points denoted by the two stacks. This is not as useful as
1949 showing the exact stack for the previous access (as
1950 --history-level=full does), but it is better than nothing, and it
1951 is almost as fast as --history-level=none.
1952 --delta-stacktrace=no|yes [default: yes on linux amd64/x86]
1954 This flag only has any effect at --history-level=full.
1955 --delta-stacktrace configures the way Helgrind captures the
1957 stacktraces for the option --history-level=full. Such a stacktrace
1958 is typically needed each time a new piece of memory is read or
1959 written in a basic block of instructions.
1960 --delta-stacktrace=no causes Helgrind to compute a full history
1962 stacktrace from the unwind info each time a stacktrace is needed.
1963 --delta-stacktrace=yes indicates to Helgrind to derive a new
1965 stacktrace from the previous stacktrace, as long as there was no
1966 call instruction, no return instruction, or any other instruction
1967 changing the call stack since the previous stacktrace was captured.
1968 If no such instruction was executed, the new stacktrace can be
1969 derived from the previous stacktrace by just changing the top frame
1970 to the current program counter. This option can speed up Helgrind
1971 by 25% when using --history-level=full.
1972 The following aspects have to be considered when using
1974 --delta-stacktrace=yes :
1975 • In some cases (for example in a function prologue), the
1977 valgrind unwinder might not properly unwind the stack, due to
1978 some limitations and/or due to wrong unwind info. When using
1979 --delta-stacktrace=yes, the wrong stack trace captured in the
1980 function prologue will be kept till the next call or return.
1981 • On the other hand, --delta-stacktrace=yes sometimes helps to
1983 obtain a correct stacktrace, for example when the unwind info
1984 allows a correct stacktrace to be done in the beginning of the
1985 sequence, but not later on in the instruction sequence.
1986 • Determining which instructions are changing the callstack is
1988 partially based on platform dependent heuristics, which have to
1989 be tuned/validated specifically for the platform. Also,
1990 unwinding in a function prologue must be good enough to allow
1991 using --delta-stacktrace=yes. Currently, the option
1992 --delta-stacktrace=yes has been reasonably validated only on
1993 linux x86 32 bits and linux amd64 64 bits. For more details
1994 about how to validate --delta-stacktrace=yes, see debug option
1995 --hg-sanity-flags and the function check_cached_rcec_ok in
1996 libhb_core.c.
1997 --conflict-cache-size=N [default: 1000000]
2000 This flag only has any effect at --history-level=full.
2001 Information about "old" conflicting accesses is stored in a cache
2003 of limited size, with LRU-style management. This is necessary
2004 because it isn't practical to store a stack trace for every single
2005 memory access made by the program. Historical information on not
2006 recently accessed locations is periodically discarded, to free up
2007 space in the cache.
2008 This option controls the size of the cache, in terms of the number
2010 of different memory addresses for which conflicting access
2011 information is stored. If you find that Helgrind is showing race
2012 errors with only one stack instead of the expected two stacks, try
2013 increasing this value.
2014 The minimum value is 10,000 and the maximum is 30,000,000 (thirty
2016 times the default value). Increasing the value by 1 increases
2017 Helgrind's memory requirement by very roughly 100 bytes, so the
2018 maximum value will easily eat up three extra gigabytes or so of
2019 memory.
2020 --check-stack-refs=no|yes [default: yes]
2022 By default Helgrind checks all data memory accesses made by your
2023 program. This flag enables you to skip checking for accesses to
2024 thread stacks (local variables). This can improve performance, but
2025 comes at the cost of missing races on stack-allocated data.
2026 --ignore-thread-creation=<yes|no> [default: no]
2028 Controls whether all activities during thread creation should be
2029 ignored. By default enabled only on Solaris. Solaris provides
2030 higher throughput, parallelism and scalability than other operating
2031 systems, at the cost of more fine-grained locking activity. This
2032 means for example that when a thread is created under glibc, just
2033 one big lock is used for all thread setup. Solaris libc uses
2034 several fine-grained locks and the creator thread resumes its
2035 activities as soon as possible, leaving for example stack and TLS
2036 setup sequence to the created thread. This situation confuses
2037 Helgrind as it assumes there is some false ordering in place
2038 between creator and created thread; and therefore many types of
2039 race conditions in the application would not be reported. To
2040 prevent such false ordering, this command line option is set to yes
2041 by default on Solaris. All activity (loads, stores, client
2042 requests) is therefore ignored during:
2043 • pthread_create() call in the creator thread
2045 • thread creation phase (stack and TLS setup) in the created
2047 thread
2048 Also new memory allocated during thread creation is untracked, that
2050 is race reporting is suppressed there. DRD does the same thing
2051 implicitly. This is necessary because Solaris libc caches many
2052 objects and reuses them for different threads and that confuses
2053 Helgrind.
2054
2056 --check-stack-var=<yes|no> [default: no]
2057 Controls whether DRD detects data races on stack variables.
2058 Verifying stack variables is disabled by default because most
2059 programs do not share stack variables over threads.
2060 --exclusive-threshold=<n> [default: off]
2062 Print an error message if any mutex or writer lock has been held
2063 longer than the time specified in milliseconds. This option enables
2064 the detection of lock contention.
2065 --join-list-vol=<n> [default: 10]
2067 Data races that occur between a statement at the end of one thread
2068 and another thread can be missed if memory access information is
2069 discarded immediately after a thread has been joined. This option
2070 allows one to specify for how many joined threads memory access
2071 information should be retained.
2072 --first-race-only=<yes|no> [default: no]
2074 Whether to report only the first data race that has been detected
2075 on a memory location or all data races that have been detected on a
2076 memory location.
2077 --free-is-write=<yes|no> [default: no]
2079 Whether to report races between accessing memory and freeing
2080 memory. Enabling this option may cause DRD to run slightly slower.
2081 Notes:
2082 • Don't enable this option when using custom memory allocators
2084 that use the VG_USERREQ__MALLOCLIKE_BLOCK and
2085 VG_USERREQ__FREELIKE_BLOCK because that would result in false
2086 positives.
2087 • Don't enable this option when using reference-counted objects
2089 because that will result in false positives, even when that
2090 code has been annotated properly with ANNOTATE_HAPPENS_BEFORE
2091 and ANNOTATE_HAPPENS_AFTER. See e.g. the output of the
2092 following command for an example: valgrind --tool=drd
2093 --free-is-write=yes drd/tests/annotate_smart_pointer.
2094 --report-signal-unlocked=<yes|no> [default: yes]
2096 Whether to report calls to pthread_cond_signal and
2097 pthread_cond_broadcast where the mutex associated with the signal
2098 through pthread_cond_wait or pthread_cond_timed_waitis not locked
2099 at the time the signal is sent. Sending a signal without holding a
2100 lock on the associated mutex is a common programming error which
2101 can cause subtle race conditions and unpredictable behavior. There
2102 exist some uncommon synchronization patterns however where it is
2103 safe to send a signal without holding a lock on the associated
2104 mutex.
2105 --segment-merging=<yes|no> [default: yes]
2107 Controls segment merging. Segment merging is an algorithm to limit
2108 memory usage of the data race detection algorithm. Disabling
2109 segment merging may improve the accuracy of the so-called 'other
2110 segments' displayed in race reports but can also trigger an out of
2111 memory error.
2112 --segment-merging-interval=<n> [default: 10]
2114 Perform segment merging only after the specified number of new
2115 segments have been created. This is an advanced configuration
2116 option that allows one to choose whether to minimize DRD's memory
2117 usage by choosing a low value or to let DRD run faster by choosing
2118 a slightly higher value. The optimal value for this parameter
2119 depends on the program being analyzed. The default value works well
2120 for most programs.
2121 --shared-threshold=<n> [default: off]
2123 Print an error message if a reader lock has been held longer than
2124 the specified time (in milliseconds). This option enables the
2125 detection of lock contention.
2126 --show-confl-seg=<yes|no> [default: yes]
2128 Show conflicting segments in race reports. Since this information
2129 can help to find the cause of a data race, this option is enabled
2130 by default. Disabling this option makes the output of DRD more
2131 compact.
2132 --show-stack-usage=<yes|no> [default: no]
2134 Print stack usage at thread exit time. When a program creates a
2135 large number of threads it becomes important to limit the amount of
2136 virtual memory allocated for thread stacks. This option makes it
2137 possible to observe how much stack memory has been used by each
2138 thread of the client program. Note: the DRD tool itself allocates
2139 some temporary data on the client thread stack. The space necessary
2140 for this temporary data must be allocated by the client program
2141 when it allocates stack memory, but is not included in stack usage
2142 reported by DRD.
2143 --ignore-thread-creation=<yes|no> [default: no]
2145 Controls whether all activities during thread creation should be
2146 ignored. By default enabled only on Solaris. Solaris provides
2147 higher throughput, parallelism and scalability than other operating
2148 systems, at the cost of more fine-grained locking activity. This
2149 means for example that when a thread is created under glibc, just
2150 one big lock is used for all thread setup. Solaris libc uses
2151 several fine-grained locks and the creator thread resumes its
2152 activities as soon as possible, leaving for example stack and TLS
2153 setup sequence to the created thread. This situation confuses DRD
2154 as it assumes there is some false ordering in place between creator
2155 and created thread; and therefore many types of race conditions in
2156 the application would not be reported. To prevent such false
2157 ordering, this command line option is set to yes by default on
2158 Solaris. All activity (loads, stores, client requests) is therefore
2159 ignored during:
2160 • pthread_create() call in the creator thread
2162 • thread creation phase (stack and TLS setup) in the created
2164 thread
2165 --trace-addr=<address> [default: none]
2167 Trace all load and store activity for the specified address. This
2168 option may be specified more than once.
2169 --ptrace-addr=<address> [default: none]
2171 Trace all load and store activity for the specified address and
2172 keep doing that even after the memory at that address has been
2173 freed and reallocated.
2174 --trace-alloc=<yes|no> [default: no]
2176 Trace all memory allocations and deallocations. May produce a huge
2177 amount of output.
2178 --trace-barrier=<yes|no> [default: no]
2180 Trace all barrier activity.
2181 --trace-cond=<yes|no> [default: no]
2183 Trace all condition variable activity.
2184 --trace-fork-join=<yes|no> [default: no]
2186 Trace all thread creation and all thread termination events.
2187 --trace-hb=<yes|no> [default: no]
2189 Trace execution of the ANNOTATE_HAPPENS_BEFORE(),
2190 ANNOTATE_HAPPENS_AFTER() and ANNOTATE_HAPPENS_DONE() client
2191 requests.
2192 --trace-mutex=<yes|no> [default: no]
2194 Trace all mutex activity.
2195 --trace-rwlock=<yes|no> [default: no]
2197 Trace all reader-writer lock activity.
2198 --trace-semaphore=<yes|no> [default: no]
2200 Trace all semaphore activity.
2201
2203 --heap=<yes|no> [default: yes]
2204 Specifies whether heap profiling should be done.
2205 --heap-admin=<size> [default: 8]
2207 If heap profiling is enabled, gives the number of administrative
2208 bytes per block to use. This should be an estimate of the average,
2209 since it may vary. For example, the allocator used by glibc on
2210 Linux requires somewhere between 4 to 15 bytes per block, depending
2211 on various factors. That allocator also requires admin space for
2212 freed blocks, but Massif cannot account for this.
2213 --stacks=<yes|no> [default: no]
2215 Specifies whether stack profiling should be done. This option slows
2216 Massif down greatly, and so is off by default. Note that Massif
2217 assumes that the main stack has size zero at start-up. This is not
2218 true, but doing otherwise accurately is difficult. Furthermore,
2219 starting at zero better indicates the size of the part of the main
2220 stack that a user program actually has control over.
2221 --pages-as-heap=<yes|no> [default: no]
2223 Tells Massif to profile memory at the page level rather than at the
2224 malloc'd block level. See above for details.
2225 --depth=<number> [default: 30]
2227 Maximum depth of the allocation trees recorded for detailed
2228 snapshots. Increasing it will make Massif run somewhat more slowly,
2229 use more memory, and produce bigger output files.
2230 --alloc-fn=<name>
2232 Functions specified with this option will be treated as though they
2233 were a heap allocation function such as malloc. This is useful for
2234 functions that are wrappers to malloc or new, which can fill up the
2235 allocation trees with uninteresting information. This option can be
2236 specified multiple times on the command line, to name multiple
2237 functions.
2238 Note that the named function will only be treated this way if it is
2240 the top entry in a stack trace, or just below another function
2241 treated this way. For example, if you have a function malloc1 that
2242 wraps malloc, and malloc2 that wraps malloc1, just specifying
2243 --alloc-fn=malloc2 will have no effect. You need to specify
2244 --alloc-fn=malloc1 as well. This is a little inconvenient, but the
2245 reason is that checking for allocation functions is slow, and it
2246 saves a lot of time if Massif can stop looking through the stack
2247 trace entries as soon as it finds one that doesn't match rather
2248 than having to continue through all the entries.
2249 Note that C++ names are demangled. Note also that overloaded C++
2251 names must be written in full. Single quotes may be necessary to
2252 prevent the shell from breaking them up. For example:
2253 --alloc-fn='operator new(unsigned, std::nothrow_t const&)'
2255 --ignore-fn=<name>
2258 Any direct heap allocation (i.e. a call to malloc, new, etc, or a
2259 call to a function named by an --alloc-fn option) that occurs in a
2260 function specified by this option will be ignored. This is mostly
2261 useful for testing purposes. This option can be specified multiple
2262 times on the command line, to name multiple functions.
2263 Any realloc of an ignored block will also be ignored, even if the
2265 realloc call does not occur in an ignored function. This avoids the
2266 possibility of negative heap sizes if ignored blocks are shrunk
2267 with realloc.
2268 The rules for writing C++ function names are the same as for
2270 --alloc-fn above.
2271 --threshold=<m.n> [default: 1.0]
2273 The significance threshold for heap allocations, as a percentage of
2274 total memory size. Allocation tree entries that account for less
2275 than this will be aggregated. Note that this should be specified in
2276 tandem with ms_print's option of the same name.
2277 --peak-inaccuracy=<m.n> [default: 1.0]
2279 Massif does not necessarily record the actual global memory
2280 allocation peak; by default it records a peak only when the global
2281 memory allocation size exceeds the previous peak by at least 1.0%.
2282 This is because there can be many local allocation peaks along the
2283 way, and doing a detailed snapshot for every one would be expensive
2284 and wasteful, as all but one of them will be later discarded. This
2285 inaccuracy can be changed (even to 0.0%) via this option, but
2286 Massif will run drastically slower as the number approaches zero.
2287 --time-unit=<i|ms|B> [default: i]
2289 The time unit used for the profiling. There are three
2290 possibilities: instructions executed (i), which is good for most
2291 cases; real (wallclock) time (ms, i.e. milliseconds), which is
2292 sometimes useful; and bytes allocated/deallocated on the heap
2293 and/or stack (B), which is useful for very short-run programs, and
2294 for testing purposes, because it is the most reproducible across
2295 different machines.
2296 --detailed-freq=<n> [default: 10]
2298 Frequency of detailed snapshots. With --detailed-freq=1, every
2299 snapshot is detailed.
2300 --max-snapshots=<n> [default: 100]
2302 The maximum number of snapshots recorded. If set to N, for all
2303 programs except very short-running ones, the final number of
2304 snapshots will be between N/2 and N.
2305 --massif-out-file=<file> [default: massif.out.%p]
2307 Write the profile data to file rather than to the default output
2308 file, massif.out.<pid>. The %p and %q format specifiers can be used
2309 to embed the process ID and/or the contents of an environment
2310 variable in the name, as is the case for the core option
2311 --log-file.
2312
2314 --bb-out-file=<name> [default: bb.out.%p]
2315 This option selects the name of the basic block vector file. The %p
2316 and %q format specifiers can be used to embed the process ID and/or
2317 the contents of an environment variable in the name, as is the case
2318 for the core option --log-file.
2319 --pc-out-file=<name> [default: pc.out.%p]
2321 This option selects the name of the PC file. This file holds
2322 program counter addresses and function name info for the various
2323 basic blocks. This can be used in conjunction with the basic block
2324 vector file to fast-forward via function names instead of just
2325 instruction counts. The %p and %q format specifiers can be used to
2326 embed the process ID and/or the contents of an environment variable
2327 in the name, as is the case for the core option --log-file.
2328 --interval-size=<number> [default: 100000000]
2330 This option selects the size of the interval to use. The default is
2331 100 million instructions, which is a commonly used value. Other
2332 sizes can be used; smaller intervals can help programs with
2333 finer-grained phases. However smaller interval size can lead to
2334 accuracy issues due to warm-up effects (When fast-forwarding the
2335 various architectural features will be un-initialized, and it will
2336 take some number of instructions before they "warm up" to the state
2337 a full simulation would be at without the fast-forwarding. Large
2338 interval sizes tend to mitigate this.)
2339 --instr-count-only [default: no]
2341 This option tells the tool to only display instruction count
2342 totals, and to not generate the actual basic block vector file.
2343 This is useful for debugging, and for gathering instruction count
2344 info without generating the large basic block vector files.
2345
2347 --basic-counts=<no|yes> [default: yes]
2348 When enabled, Lackey prints the following statistics and
2349 information about the execution of the client program:
2350 1. The number of calls to the function specified by the --fnname
2352 option (the default is main). If the program has had its
2353 symbols stripped, the count will always be zero.
2354 2. The number of conditional branches encountered and the number
2356 and proportion of those taken.
2357 3. The number of superblocks entered and completed by the program.
2359 Note that due to optimisations done by the JIT, this is not at
2360 all an accurate value.
2361 4. The number of guest (x86, amd64, ppc, etc.) instructions and IR
2363 statements executed. IR is Valgrind's RISC-like intermediate
2364 representation via which all instrumentation is done.
2365 5. Ratios between some of these counts.
2367 6. The exit code of the client program.
2369 --detailed-counts=<no|yes> [default: no]
2371 When enabled, Lackey prints a table containing counts of loads,
2372 stores and ALU operations, differentiated by their IR types. The IR
2373 types are identified by their IR name ("I1", "I8", ... "I128",
2374 "F32", "F64", and "V128").
2375 --trace-mem=<no|yes> [default: no]
2377 When enabled, Lackey prints the size and address of almost every
2378 memory access made by the program. See the comments at the top of
2379 the file lackey/lk_main.c for details about the output format, how
2380 it works, and inaccuracies in the address trace. Note that this
2381 option produces immense amounts of output.
2382 --trace-superblocks=<no|yes> [default: no]
2384 When enabled, Lackey prints out the address of every superblock (a
2385 single entry, multiple exit, linear chunk of code) executed by the
2386 program. This is primarily of interest to Valgrind developers. See
2387 the comments at the top of the file lackey/lk_main.c for details
2388 about the output format. Note that this option produces large
2389 amounts of output.
2390 --fnname=<name> [default: main]
2392 Changes the function for which calls are counted when
2393 --basic-counts=yes is specified.
2394
2396 Valgrind supports the downloading of debuginfo files via debuginfod, an
2397 HTTP server for distributing ELF/DWARF debugging information. When a
2398 debuginfo file cannot be found locally, Valgrind is able to query
2399 debuginfod servers for the file using its build-id.
2400 In order to use this feature debuginfod-find must be installed and
2402 $DEBUGINFOD_URLS must contain URLs of debuginfod servers. Valgrind does
2403 not support debuginfod-find verbose output that is normally enabled
2404 with $DEBUGINFOD_PROGRESS and $DEBUGINFOD_VERBOSE. These environment
2405 variables will be ignored.
2406 For more information regarding debuginfod, see
2408 https://sourceware.org/elfutils/Debuginfod.html
2409
2411 cg_annotate(1), callgrind_annotate(1), callgrind_control(1),
2412 ms_print(1), $INSTALL/share/doc/valgrind/html/index.html or
2413 http://www.valgrind.org/docs/manual/index.html, Debugging your program
2414 using Valgrind's gdbserver and GDB[1] vgdb[2], Valgrind monitor
2415 commands[3], The Commentary[4], Scheduling and Multi-Thread
2416 Performance[5], Cachegrind: a cache and branch-prediction profiler[6].
2417 Execution Trees[7]
2418
2420 See the AUTHORS file in the valgrind distribution for a comprehensive
2421 list of authors.
2422 This manpage was written by Andres Roldan <aroldan@debian.org> and the
2424 Valgrind developers.
2425
2427 1. Debugging your program using Valgrind's gdbserver and GDB
2428 http://www.valgrind.org/docs/manual/manual-core-adv.html#manual-core-adv.gdbserver
2429 2. vgdb
2431 http://www.valgrind.org/docs/manual/manual-core-adv.html#manual-core-adv.vgdb
2432 3. Valgrind monitor commands
2434 http://www.valgrind.org/docs/manual/manual-core-adv.html#manual-core-adv.valgrind-monitor-commands
2435 4. The Commentary
2437 http://www.valgrind.org/docs/manual/manual-core.html#manual-core.comment
2438 5. Scheduling and Multi-Thread Performance
2440 http://www.valgrind.org/docs/manual/manual-core.html#manual-core.pthreads_perf_sched
2441 6. Cachegrind: a cache and branch-prediction profiler
2443 http://www.valgrind.org/docs/manual/cg-manual.html
2444 7. Execution Trees
2446 http://www.valgrind.org/docs/manual/manual-core.html#manual-core.xtree
2447Release 3.17.0 03/19/2021 VALGRIND(1)