1VALGRIND(1) Release 3.15.0 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-sgcheck,
33 exp-bbv, etc.
34
36 These options work with all tools.
37 -h --help
39 Show help for all options, both for the core and for the selected
40 tool. If the option is repeated it is equivalent to giving
41 --help-debug.
42 --help-debug
44 Same as --help, but also lists debugging options which usually are
45 only of use to Valgrind's developers.
46 --version
48 Show the version number of the Valgrind core. Tools can have their
49 own version numbers. There is a scheme in place to ensure that
50 tools only execute when the core version is one they are known to
51 work with. This was done to minimise the chances of strange
52 problems arising from tool-vs-core version incompatibilities.
53 -q, --quiet
55 Run silently, and only print error messages. Useful if you are
56 running regression tests or have some other automated test
57 machinery.
58 -v, --verbose
60 Be more verbose. Gives extra information on various aspects of your
61 program, such as: the shared objects loaded, the suppressions used,
62 the progress of the instrumentation and execution engines, and
63 warnings about unusual behaviour. Repeating the option increases
64 the verbosity level.
65 --trace-children=<yes|no> [default: no]
67 When enabled, Valgrind will trace into sub-processes initiated via
68 the exec system call. This is necessary for multi-process programs.
69 Note that Valgrind does trace into the child of a fork (it would be
71 difficult not to, since fork makes an identical copy of a process),
72 so this option is arguably badly named. However, most children of
73 fork calls immediately call exec anyway.
74 --trace-children-skip=patt1,patt2,...
76 This option only has an effect when --trace-children=yes is
77 specified. It allows for some children to be skipped. The option
78 takes a comma separated list of patterns for the names of child
79 executables that Valgrind should not trace into. Patterns may
80 include the metacharacters ? and *, which have the usual meaning.
81 This can be useful for pruning uninteresting branches from a tree
83 of processes being run on Valgrind. But you should be careful when
84 using it. When Valgrind skips tracing into an executable, it
85 doesn't just skip tracing that executable, it also skips tracing
86 any of that executable's child processes. In other words, the flag
87 doesn't merely cause tracing to stop at the specified executables
88 -- it skips tracing of entire process subtrees rooted at any of the
89 specified executables.
90 --trace-children-skip-by-arg=patt1,patt2,...
92 This is the same as --trace-children-skip, with one difference: the
93 decision as to whether to trace into a child process is made by
94 examining the arguments to the child process, rather than the name
95 of its executable.
96 --child-silent-after-fork=<yes|no> [default: no]
98 When enabled, Valgrind will not show any debugging or logging
99 output for the child process resulting from a fork call. This can
100 make the output less confusing (although more misleading) when
101 dealing with processes that create children. It is particularly
102 useful in conjunction with --trace-children=. Use of this option is
103 also strongly recommended if you are requesting XML output
104 (--xml=yes), since otherwise the XML from child and parent may
105 become mixed up, which usually makes it useless.
106 --vgdb=<no|yes|full> [default: yes]
108 Valgrind will provide "gdbserver" functionality when --vgdb=yes or
109 --vgdb=full is specified. This allows an external GNU GDB debugger
110 to control and debug your program when it runs on Valgrind.
111 --vgdb=full incurs significant performance overheads, but provides
112 more precise breakpoints and watchpoints. See Debugging your
113 program using Valgrind's gdbserver and GDB for a detailed
114 description.
115 If the embedded gdbserver is enabled but no gdb is currently being
117 used, the vgdb command line utility can send "monitor commands" to
118 Valgrind from a shell. The Valgrind core provides a set of Valgrind
119 monitor commands. A tool can optionally provide tool specific
120 monitor commands, which are documented in the tool specific
121 chapter.
122 --vgdb-error=<number> [default: 999999999]
124 Use this option when the Valgrind gdbserver is enabled with
125 --vgdb=yes or --vgdb=full. Tools that report errors will wait for
126 "number" errors to be reported before freezing the program and
127 waiting for you to connect with GDB. It follows that a value of
128 zero will cause the gdbserver to be started before your program is
129 executed. This is typically used to insert GDB breakpoints before
130 execution, and also works with tools that do not report errors,
131 such as Massif.
132 --vgdb-stop-at=<set> [default: none]
134 Use this option when the Valgrind gdbserver is enabled with
135 --vgdb=yes or --vgdb=full. The Valgrind gdbserver will be invoked
136 for each error after --vgdb-error have been reported. You can
137 additionally ask the Valgrind gdbserver to be invoked for other
138 events, specified in one of the following ways:
139 · a comma separated list of one or more of startup exit
141 valgrindabexit.
142 The values startup exit valgrindabexit respectively indicate to
144 invoke gdbserver before your program is executed, after the
145 last instruction of your program, on Valgrind abnormal exit
146 (e.g. internal error, out of memory, ...).
147 Note: startup and --vgdb-error=0 will both cause Valgrind
149 gdbserver to be invoked before your program is executed. The
150 --vgdb-error=0 will in addition cause your program to stop on
151 all subsequent errors.
152 · all to specify the complete set. It is equivalent to
154 --vgdb-stop-at=startup,exit,valgrindabexit.
155 · none for the empty set.
157 --track-fds=<yes|no> [default: no]
159 When enabled, Valgrind will print out a list of open file
160 descriptors on exit or on request, via the gdbserver monitor
161 command v.info open_fds. Along with each file descriptor is printed
162 a stack backtrace of where the file was opened and any details
163 relating to the file descriptor such as the file name or socket
164 details.
165 --time-stamp=<yes|no> [default: no]
167 When enabled, each message is preceded with an indication of the
168 elapsed wallclock time since startup, expressed as days, hours,
169 minutes, seconds and milliseconds.
170 --log-fd=<number> [default: 2, stderr]
172 Specifies that Valgrind should send all of its messages to the
173 specified file descriptor. The default, 2, is the standard error
174 channel (stderr). Note that this may interfere with the client's
175 own use of stderr, as Valgrind's output will be interleaved with
176 any output that the client sends to stderr.
177 --log-file=<filename>
179 Specifies that Valgrind should send all of its messages to the
180 specified file. If the file name is empty, it causes an abort.
181 There are three special format specifiers that can be used in the
182 file name.
183 %p is replaced with the current process ID. This is very useful for
185 program that invoke multiple processes. WARNING: If you use
186 --trace-children=yes and your program invokes multiple processes OR
187 your program forks without calling exec afterwards, and you don't
188 use this specifier (or the %q specifier below), the Valgrind output
189 from all those processes will go into one file, possibly jumbled
190 up, and possibly incomplete. Note: If the program forks and calls
191 exec afterwards, Valgrind output of the child from the period
192 between fork and exec will be lost. Fortunately this gap is really
193 tiny for most programs; and modern programs use posix_spawn anyway.
194 %n is replaced with a file sequence number unique for this process.
196 This is useful for processes that produces several files from the
197 same filename template.
198 %q{FOO} is replaced with the contents of the environment variable
200 FOO. If the {FOO} part is malformed, it causes an abort. This
201 specifier is rarely needed, but very useful in certain
202 circumstances (eg. when running MPI programs). The idea is that you
203 specify a variable which will be set differently for each process
204 in the job, for example BPROC_RANK or whatever is applicable in
205 your MPI setup. If the named environment variable is not set, it
206 causes an abort. Note that in some shells, the { and } characters
207 may need to be escaped with a backslash.
208 %% is replaced with %.
210 If an % is followed by any other character, it causes an abort.
212 If the file name specifies a relative file name, it is put in the
214 program's initial working directory: this is the current directory
215 when the program started its execution after the fork or after the
216 exec. If it specifies an absolute file name (ie. starts with '/')
217 then it is put there.
218 --log-socket=<ip-address:port-number>
220 Specifies that Valgrind should send all of its messages to the
221 specified port at the specified IP address. The port may be
222 omitted, in which case port 1500 is used. If a connection cannot be
223 made to the specified socket, Valgrind falls back to writing output
224 to the standard error (stderr). This option is intended to be used
225 in conjunction with the valgrind-listener program. For further
226 details, see the commentary in the manual.
227
229 These options are used by all tools that can report errors, e.g.
230 Memcheck, but not Cachegrind.
231 --xml=<yes|no> [default: no]
233 When enabled, the important parts of the output (e.g. tool error
234 messages) will be in XML format rather than plain text.
235 Furthermore, the XML output will be sent to a different output
236 channel than the plain text output. Therefore, you also must use
237 one of --xml-fd, --xml-file or --xml-socket to specify where the
238 XML is to be sent.
239 Less important messages will still be printed in plain text, but
241 because the XML output and plain text output are sent to different
242 output channels (the destination of the plain text output is still
243 controlled by --log-fd, --log-file and --log-socket) this should
244 not cause problems.
245 This option is aimed at making life easier for tools that consume
247 Valgrind's output as input, such as GUI front ends. Currently this
248 option works with Memcheck, Helgrind, DRD and SGcheck. The output
249 format is specified in the file
250 docs/internals/xml-output-protocol4.txt in the source tree for
251 Valgrind 3.5.0 or later.
252 The recommended options for a GUI to pass, when requesting XML
254 output, are: --xml=yes to enable XML output, --xml-file to send the
255 XML output to a (presumably GUI-selected) file, --log-file to send
256 the plain text output to a second GUI-selected file,
257 --child-silent-after-fork=yes, and -q to restrict the plain text
258 output to critical error messages created by Valgrind itself. For
259 example, failure to read a specified suppressions file counts as a
260 critical error message. In this way, for a successful run the text
261 output file will be empty. But if it isn't empty, then it will
262 contain important information which the GUI user should be made
263 aware of.
264 --xml-fd=<number> [default: -1, disabled]
266 Specifies that Valgrind should send its XML output to the specified
267 file descriptor. It must be used in conjunction with --xml=yes.
268 --xml-file=<filename>
270 Specifies that Valgrind should send its XML output to the specified
271 file. It must be used in conjunction with --xml=yes. Any %p or %q
272 sequences appearing in the filename are expanded in exactly the
273 same way as they are for --log-file. See the description of --log-
274 file for details.
275 --xml-socket=<ip-address:port-number>
277 Specifies that Valgrind should send its XML output the specified
278 port at the specified IP address. It must be used in conjunction
279 with --xml=yes. The form of the argument is the same as that used
280 by --log-socket. See the description of --log-socket for further
281 details.
282 --xml-user-comment=<string>
284 Embeds an extra user comment string at the start of the XML output.
285 Only works when --xml=yes is specified; ignored otherwise.
286 --demangle=<yes|no> [default: yes]
288 Enable/disable automatic demangling (decoding) of C++ names.
289 Enabled by default. When enabled, Valgrind will attempt to
290 translate encoded C++ names back to something approaching the
291 original. The demangler handles symbols mangled by g++ versions
292 2.X, 3.X and 4.X.
293 An important fact about demangling is that function names mentioned
295 in suppressions files should be in their mangled form. Valgrind
296 does not demangle function names when searching for applicable
297 suppressions, because to do otherwise would make suppression file
298 contents dependent on the state of Valgrind's demangling machinery,
299 and also slow down suppression matching.
300 --num-callers=<number> [default: 12]
302 Specifies the maximum number of entries shown in stack traces that
303 identify program locations. Note that errors are commoned up using
304 only the top four function locations (the place in the current
305 function, and that of its three immediate callers). So this doesn't
306 affect the total number of errors reported.
307 The maximum value for this is 500. Note that higher settings will
309 make Valgrind run a bit more slowly and take a bit more memory, but
310 can be useful when working with programs with deeply-nested call
311 chains.
312 --unw-stack-scan-thresh=<number> [default: 0] ,
314 --unw-stack-scan-frames=<number> [default: 5]
315 Stack-scanning support is available only on ARM targets.
316 These flags enable and control stack unwinding by stack scanning.
318 When the normal stack unwinding mechanisms -- usage of Dwarf CFI
319 records, and frame-pointer following -- fail, stack scanning may be
320 able to recover a stack trace.
321 Note that stack scanning is an imprecise, heuristic mechanism that
323 may give very misleading results, or none at all. It should be used
324 only in emergencies, when normal unwinding fails, and it is
325 important to nevertheless have stack traces.
326 Stack scanning is a simple technique: the unwinder reads words from
328 the stack, and tries to guess which of them might be return
329 addresses, by checking to see if they point just after ARM or Thumb
330 call instructions. If so, the word is added to the backtrace.
331 The main danger occurs when a function call returns, leaving its
333 return address exposed, and a new function is called, but the new
334 function does not overwrite the old address. The result of this is
335 that the backtrace may contain entries for functions which have
336 already returned, and so be very confusing.
337 A second limitation of this implementation is that it will scan
339 only the page (4KB, normally) containing the starting stack
340 pointer. If the stack frames are large, this may result in only a
341 few (or not even any) being present in the trace. Also, if you are
342 unlucky and have an initial stack pointer near the end of its
343 containing page, the scan may miss all interesting frames.
344 By default stack scanning is disabled. The normal use case is to
346 ask for it when a stack trace would otherwise be very short. So, to
347 enable it, use --unw-stack-scan-thresh=number. This requests
348 Valgrind to try using stack scanning to "extend" stack traces which
349 contain fewer than number frames.
350 If stack scanning does take place, it will only generate at most
352 the number of frames specified by --unw-stack-scan-frames.
353 Typically, stack scanning generates so many garbage entries that
354 this value is set to a low value (5) by default. In no case will a
355 stack trace larger than the value specified by --num-callers be
356 created.
357 --error-limit=<yes|no> [default: yes]
359 When enabled, Valgrind stops reporting errors after 10,000,000 in
360 total, or 1,000 different ones, have been seen. This is to stop the
361 error tracking machinery from becoming a huge performance overhead
362 in programs with many errors.
363 --error-exitcode=<number> [default: 0]
365 Specifies an alternative exit code to return if Valgrind reported
366 any errors in the run. When set to the default value (zero), the
367 return value from Valgrind will always be the return value of the
368 process being simulated. When set to a nonzero value, that value is
369 returned instead, if Valgrind detects any errors. This is useful
370 for using Valgrind as part of an automated test suite, since it
371 makes it easy to detect test cases for which Valgrind has reported
372 errors, just by inspecting return codes.
373 --exit-on-first-error=<yes|no> [default: no]
375 If this option is enabled, Valgrind exits on the first error. A
376 nonzero exit value must be defined using --error-exitcode option.
377 Useful if you are running regression tests or have some other
378 automated test machinery.
379 --error-markers=<begin>,<end> [default: none]
381 When errors are output as plain text (i.e. XML not used),
382 --error-markers instructs to output a line containing the begin
383 (end) string before (after) each error.
384 Such marker lines facilitate searching for errors and/or extracting
386 errors in an output file that contain valgrind errors mixed with
387 the program output.
388 Note that empty markers are accepted. So, only using a begin (or an
390 end) marker is possible.
391 --show-error-list=no|yes [default: no]
393 If this option is enabled, for tools that report errors, valgrind
394 will show the list of detected errors and the list of used
395 suppressions at exit.
396 Note that at verbosity 2 and above, valgrind automatically shows
398 the list of detected errors and the list of used suppressions at
399 exit, unless --show-error-list=no is selected.
400 -s
402 Specifying -s is equivalent to --show-error-list=yes.
403 --sigill-diagnostics=<yes|no> [default: yes]
405 Enable/disable printing of illegal instruction diagnostics. Enabled
406 by default, but defaults to disabled when --quiet is given. The
407 default can always be explicitly overridden by giving this option.
408 When enabled, a warning message will be printed, along with some
410 diagnostics, whenever an instruction is encountered that Valgrind
411 cannot decode or translate, before the program is given a SIGILL
412 signal. Often an illegal instruction indicates a bug in the program
413 or missing support for the particular instruction in Valgrind. But
414 some programs do deliberately try to execute an instruction that
415 might be missing and trap the SIGILL signal to detect processor
416 features. Using this flag makes it possible to avoid the diagnostic
417 output that you would otherwise get in such cases.
418 --keep-debuginfo=<yes|no> [default: no]
420 When enabled, keep ("archive") symbols and all other debuginfo for
421 unloaded code. This allows saved stack traces to include file/line
422 info for code that has been dlclose'd (or similar). Be careful with
423 this, since it can lead to unbounded memory use for programs which
424 repeatedly load and unload shared objects.
425 Some tools and some functionalities have only limited support for
427 archived debug info. Memcheck fully supports it. Generally, tools
428 that report errors can use archived debug info to show the error
429 stack traces. The known limitations are: Helgrind's past access
430 stack trace of a race condition is does not use archived debug
431 info. Massif (and more generally the xtree Massif output format)
432 does not make use of archived debug info. Only Memcheck has been
433 (somewhat) tested with --keep-debuginfo=yes, so other tools may
434 have unknown limitations.
435 --show-below-main=<yes|no> [default: no]
437 By default, stack traces for errors do not show any functions that
438 appear beneath main because most of the time it's uninteresting C
439 library stuff and/or gobbledygook. Alternatively, if main is not
440 present in the stack trace, stack traces will not show any
441 functions below main-like functions such as glibc's
442 __libc_start_main. Furthermore, if main-like functions are present
443 in the trace, they are normalised as (below main), in order to make
444 the output more deterministic.
445 If this option is enabled, all stack trace entries will be shown
447 and main-like functions will not be normalised.
448 --fullpath-after=<string> [default: don't show source paths]
450 By default Valgrind only shows the filenames in stack traces, but
451 not full paths to source files. When using Valgrind in large
452 projects where the sources reside in multiple different
453 directories, this can be inconvenient. --fullpath-after provides a
454 flexible solution to this problem. When this option is present, the
455 path to each source file is shown, with the following all-important
456 caveat: if string is found in the path, then the path up to and
457 including string is omitted, else the path is shown unmodified.
458 Note that string is not required to be a prefix of the path.
459 For example, consider a file named
461 /home/janedoe/blah/src/foo/bar/xyzzy.c. Specifying
462 --fullpath-after=/home/janedoe/blah/src/ will cause Valgrind to
463 show the name as foo/bar/xyzzy.c.
464 Because the string is not required to be a prefix,
466 --fullpath-after=src/ will produce the same output. This is useful
467 when the path contains arbitrary machine-generated characters. For
468 example, the path /my/build/dir/C32A1B47/blah/src/foo/xyzzy can be
469 pruned to foo/xyzzy using --fullpath-after=/blah/src/.
470 If you simply want to see the full path, just specify an empty
472 string: --fullpath-after=. This isn't a special case, merely a
473 logical consequence of the above rules.
474 Finally, you can use --fullpath-after multiple times. Any
476 appearance of it causes Valgrind to switch to producing full paths
477 and applying the above filtering rule. Each produced path is
478 compared against all the --fullpath-after-specified strings, in the
479 order specified. The first string to match causes the path to be
480 truncated as described above. If none match, the full path is
481 shown. This facilitates chopping off prefixes when the sources are
482 drawn from a number of unrelated directories.
483 --extra-debuginfo-path=<path> [default: undefined and unused]
485 By default Valgrind searches in several well-known paths for debug
486 objects, such as /usr/lib/debug/.
487 However, there may be scenarios where you may wish to put debug
489 objects at an arbitrary location, such as external storage when
490 running Valgrind on a mobile device with limited local storage.
491 Another example might be a situation where you do not have
492 permission to install debug object packages on the system where you
493 are running Valgrind.
494 In these scenarios, you may provide an absolute path as an extra,
496 final place for Valgrind to search for debug objects by specifying
497 --extra-debuginfo-path=/path/to/debug/objects. The given path will
498 be prepended to the absolute path name of the searched-for object.
499 For example, if Valgrind is looking for the debuginfo for
500 /w/x/y/zz.so and --extra-debuginfo-path=/a/b/c is specified, it
501 will look for a debug object at /a/b/c/w/x/y/zz.so.
502 This flag should only be specified once. If it is specified
504 multiple times, only the last instance is honoured.
505 --debuginfo-server=ipaddr:port [default: undefined and unused]
507 This is a new, experimental, feature introduced in version 3.9.0.
508 In some scenarios it may be convenient to read debuginfo from
510 objects stored on a different machine. With this flag, Valgrind
511 will query a debuginfo server running on ipaddr and listening on
512 port port, if it cannot find the debuginfo object in the local
513 filesystem.
514 The debuginfo server must accept TCP connections on port port. The
516 debuginfo server is contained in the source file
517 auxprogs/valgrind-di-server.c. It will only serve from the
518 directory it is started in. port defaults to 1500 in both client
519 and server if not specified.
520 If Valgrind looks for the debuginfo for /w/x/y/zz.so by using the
522 debuginfo server, it will strip the pathname components and merely
523 request zz.so on the server. That in turn will look only in its
524 current working directory for a matching debuginfo object.
525 The debuginfo data is transmitted in small fragments (8 KB) as
527 requested by Valgrind. Each block is compressed using LZO to reduce
528 transmission time. The implementation has been tuned for best
529 performance over a single-stage 802.11g (WiFi) network link.
530 Note that checks for matching primary vs debug objects, using GNU
532 debuglink CRC scheme, are performed even when using the debuginfo
533 server. To disable such checking, you need to also specify
534 --allow-mismatched-debuginfo=yes.
535 By default the Valgrind build system will build valgrind-di-server
537 for the target platform, which is almost certainly not what you
538 want. So far we have been unable to find out how to get
539 automake/autoconf to build it for the build platform. If you want
540 to use it, you will have to recompile it by hand using the command
541 shown at the top of auxprogs/valgrind-di-server.c.
542 --allow-mismatched-debuginfo=no|yes [no]
544 When reading debuginfo from separate debuginfo objects, Valgrind
545 will by default check that the main and debuginfo objects match,
546 using the GNU debuglink mechanism. This guarantees that it does not
547 read debuginfo from out of date debuginfo objects, and also ensures
548 that Valgrind can't crash as a result of mismatches.
549 This check can be overridden using
551 --allow-mismatched-debuginfo=yes. This may be useful when the
552 debuginfo and main objects have not been split in the proper way.
553 Be careful when using this, though: it disables all consistency
554 checking, and Valgrind has been observed to crash when the main and
555 debuginfo objects don't match.
556 --suppressions=<filename> [default: $PREFIX/lib/valgrind/default.supp]
558 Specifies an extra file from which to read descriptions of errors
559 to suppress. You may use up to 100 extra suppression files.
560 --gen-suppressions=<yes|no|all> [default: no]
562 When set to yes, Valgrind will pause after every error shown and
563 print the line:
564 ---- Print suppression ? --- [Return/N/n/Y/y/C/c] ----
566 Pressing Ret, or N Ret or n Ret, causes Valgrind continue execution
568 without printing a suppression for this error.
569 Pressing Y Ret or y Ret causes Valgrind to write a suppression for
571 this error. You can then cut and paste it into a suppression file
572 if you don't want to hear about the error in the future.
573 When set to all, Valgrind will print a suppression for every
575 reported error, without querying the user.
576 This option is particularly useful with C++ programs, as it prints
578 out the suppressions with mangled names, as required.
579 Note that the suppressions printed are as specific as possible. You
581 may want to common up similar ones, by adding wildcards to function
582 names, and by using frame-level wildcards. The wildcarding
583 facilities are powerful yet flexible, and with a bit of careful
584 editing, you may be able to suppress a whole family of related
585 errors with only a few suppressions.
586 Sometimes two different errors are suppressed by the same
588 suppression, in which case Valgrind will output the suppression
589 more than once, but you only need to have one copy in your
590 suppression file (but having more than one won't cause problems).
591 Also, the suppression name is given as <insert a suppression name
592 here>; the name doesn't really matter, it's only used with the -v
593 option which prints out all used suppression records.
594 --input-fd=<number> [default: 0, stdin]
596 When using --gen-suppressions=yes, Valgrind will stop so as to read
597 keyboard input from you when each error occurs. By default it reads
598 from the standard input (stdin), which is problematic for programs
599 which close stdin. This option allows you to specify an alternative
600 file descriptor from which to read input.
601 --dsymutil=no|yes [yes]
603 This option is only relevant when running Valgrind on Mac OS X.
604 Mac OS X uses a deferred debug information (debuginfo) linking
606 scheme. When object files containing debuginfo are linked into a
607 .dylib or an executable, the debuginfo is not copied into the final
608 file. Instead, the debuginfo must be linked manually by running
609 dsymutil, a system-provided utility, on the executable or .dylib.
610 The resulting combined debuginfo is placed in a directory alongside
611 the executable or .dylib, but with the extension .dSYM.
612 With --dsymutil=no, Valgrind will detect cases where the .dSYM
614 directory is either missing, or is present but does not appear to
615 match the associated executable or .dylib, most likely because it
616 is out of date. In these cases, Valgrind will print a warning
617 message but take no further action.
618 With --dsymutil=yes, Valgrind will, in such cases, automatically
620 run dsymutil as necessary to bring the debuginfo up to date. For
621 all practical purposes, if you always use --dsymutil=yes, then
622 there is never any need to run dsymutil manually or as part of your
623 applications's build system, since Valgrind will run it as
624 necessary.
625 Valgrind will not attempt to run dsymutil on any executable or
627 library in /usr/, /bin/, /sbin/, /opt/, /sw/, /System/, /Library/
628 or /Applications/ since dsymutil will always fail in such
629 situations. It fails both because the debuginfo for such
630 pre-installed system components is not available anywhere, and also
631 because it would require write privileges in those directories.
632 Be careful when using --dsymutil=yes, since it will cause
634 pre-existing .dSYM directories to be silently deleted and
635 re-created. Also note that dsymutil is quite slow, sometimes
636 excessively so.
637 --max-stackframe=<number> [default: 2000000]
639 The maximum size of a stack frame. If the stack pointer moves by
640 more than this amount then Valgrind will assume that the program is
641 switching to a different stack.
642 You may need to use this option if your program has large
644 stack-allocated arrays. Valgrind keeps track of your program's
645 stack pointer. If it changes by more than the threshold amount,
646 Valgrind assumes your program is switching to a different stack,
647 and Memcheck behaves differently than it would for a stack pointer
648 change smaller than the threshold. Usually this heuristic works
649 well. However, if your program allocates large structures on the
650 stack, this heuristic will be fooled, and Memcheck will
651 subsequently report large numbers of invalid stack accesses. This
652 option allows you to change the threshold to a different value.
653 You should only consider use of this option if Valgrind's debug
655 output directs you to do so. In that case it will tell you the new
656 threshold you should specify.
657 In general, allocating large structures on the stack is a bad idea,
659 because you can easily run out of stack space, especially on
660 systems with limited memory or which expect to support large
661 numbers of threads each with a small stack, and also because the
662 error checking performed by Memcheck is more effective for
663 heap-allocated data than for stack-allocated data. If you have to
664 use this option, you may wish to consider rewriting your code to
665 allocate on the heap rather than on the stack.
666 --main-stacksize=<number> [default: use current 'ulimit' value]
668 Specifies the size of the main thread's stack.
669 To simplify its memory management, Valgrind reserves all required
671 space for the main thread's stack at startup. That means it needs
672 to know the required stack size at startup.
673 By default, Valgrind uses the current "ulimit" value for the stack
675 size, or 16 MB, whichever is lower. In many cases this gives a
676 stack size in the range 8 to 16 MB, which almost never overflows
677 for most applications.
678 If you need a larger total stack size, use --main-stacksize to
680 specify it. Only set it as high as you need, since reserving far
681 more space than you need (that is, hundreds of megabytes more than
682 you need) constrains Valgrind's memory allocators and may reduce
683 the total amount of memory that Valgrind can use. This is only
684 really of significance on 32-bit machines.
685 On Linux, you may request a stack of size up to 2GB. Valgrind will
687 stop with a diagnostic message if the stack cannot be allocated.
688 --main-stacksize only affects the stack size for the program's
690 initial thread. It has no bearing on the size of thread stacks, as
691 Valgrind does not allocate those.
692 You may need to use both --main-stacksize and --max-stackframe
694 together. It is important to understand that --main-stacksize sets
695 the maximum total stack size, whilst --max-stackframe specifies the
696 largest size of any one stack frame. You will have to work out the
697 --main-stacksize value for yourself (usually, if your applications
698 segfaults). But Valgrind will tell you the needed --max-stackframe
699 size, if necessary.
700 As discussed further in the description of --max-stackframe, a
702 requirement for a large stack is a sign of potential portability
703 problems. You are best advised to place all large data in
704 heap-allocated memory.
705 --max-threads=<number> [default: 500]
707 By default, Valgrind can handle to up to 500 threads. Occasionally,
708 that number is too small. Use this option to provide a different
709 limit. E.g. --max-threads=3000.
710
712 For tools that use their own version of malloc (e.g. Memcheck, Massif,
713 Helgrind, DRD), the following options apply.
714 --alignment=<number> [default: 8 or 16, depending on the platform]
716 By default Valgrind's malloc, realloc, etc, return a block whose
717 starting address is 8-byte aligned or 16-byte aligned (the value
718 depends on the platform and matches the platform default). This
719 option allows you to specify a different alignment. The supplied
720 value must be greater than or equal to the default, less than or
721 equal to 4096, and must be a power of two.
722 --redzone-size=<number> [default: depends on the tool]
724 Valgrind's malloc, realloc, etc, add padding blocks before and
725 after each heap block allocated by the program being run. Such
726 padding blocks are called redzones. The default value for the
727 redzone size depends on the tool. For example, Memcheck adds and
728 protects a minimum of 16 bytes before and after each block
729 allocated by the client. This allows it to detect block underruns
730 or overruns of up to 16 bytes.
731 Increasing the redzone size makes it possible to detect overruns of
733 larger distances, but increases the amount of memory used by
734 Valgrind. Decreasing the redzone size will reduce the memory needed
735 by Valgrind but also reduces the chances of detecting
736 over/underruns, so is not recommended.
737 --xtree-memory=none|allocs|full [none]
739 Tools replacing Valgrind's malloc, realloc, etc, can optionally
740 produce an execution tree detailing which piece of code is
741 responsible for heap memory usage. See ??? for a detailed
742 explanation about execution trees.
743 When set to none, no memory execution tree is produced.
745 When set to allocs, the memory execution tree gives the current
747 number of allocated bytes and the current number of allocated
748 blocks.
749 When set to full, the memory execution tree gives 6 different
751 measurements : the current number of allocated bytes and blocks
752 (same values as for allocs), the total number of allocated bytes
753 and blocks, the total number of freed bytes and blocks.
754 Note that the overhead in cpu and memory to produce an xtree
756 depends on the tool. The overhead in cpu is small for the value
757 allocs, as the information needed to produce this report is
758 maintained in any case by the tool. For massif and helgrind,
759 specifying full implies to capture a stack trace for each free
760 operation, while normally these tools only capture an allocation
761 stack trace. For Memcheck, the cpu overhead for the value full is
762 small, as this can only be used in combination with
763 --keep-stacktraces=alloc-and-free or
764 --keep-stacktraces=alloc-then-free, which already records a stack
765 trace for each free operation. The memory overhead varies between 5
766 and 10 words per unique stacktrace in the xtree, plus the memory
767 needed to record the stack trace for the free operations, if needed
768 specifically for the xtree.
769 --xtree-memory-file=<filename> [default: xtmemory.kcg.%p]
771 Specifies that Valgrind should produce the xtree memory report in
772 the specified file. Any %p or %q sequences appearing in the
773 filename are expanded in exactly the same way as they are for
774 --log-file. See the description of --log-file for details.
775 If the filename contains the extension .ms, then the produced file
777 format will be a massif output file format. If the filename
778 contains the extension .kcg or no extension is provided or
779 recognised, then the produced file format will be a callgrind
780 output format.
781 See ??? for a detailed explanation about execution trees formats.
783
785 These options apply to all tools, as they affect certain obscure
786 workings of the Valgrind core. Most people won't need to use them.
787 --smc-check=<none|stack|all|all-non-file> [default: all-non-file for
789 x86/amd64/s390x, stack for other archs]
790 This option controls Valgrind's detection of self-modifying code.
791 If no checking is done, when a program executes some code, then
792 overwrites it with new code, and executes the new code, Valgrind
793 will continue to execute the translations it made for the old code.
794 This will likely lead to incorrect behaviour and/or crashes.
795 For "modern" architectures -- anything that's not x86, amd64 or
797 s390x -- the default is stack. This is because a correct program
798 must take explicit action to reestablish D-I cache coherence
799 following code modification. Valgrind observes and honours such
800 actions, with the result that self-modifying code is transparently
801 handled with zero extra cost.
802 For x86, amd64 and s390x, the program is not required to notify the
804 hardware of required D-I coherence syncing. Hence the default is
805 all-non-file, which covers the normal case of generating code into
806 an anonymous (non-file-backed) mmap'd area.
807 The meanings of the four available settings are as follows. No
809 detection (none), detect self-modifying code on the stack (which is
810 used by GCC to implement nested functions) (stack), detect
811 self-modifying code everywhere (all), and detect self-modifying
812 code everywhere except in file-backed mappings (all-non-file).
813 Running with all will slow Valgrind down noticeably. Running with
815 none will rarely speed things up, since very little code gets
816 dynamically generated in most programs. The
817 VALGRIND_DISCARD_TRANSLATIONS client request is an alternative to
818 --smc-check=all and --smc-check=all-non-file that requires more
819 programmer effort but allows Valgrind to run your program faster,
820 by telling it precisely when translations need to be re-made.
821 --smc-check=all-non-file provides a cheaper but more limited
823 version of --smc-check=all. It adds checks to any translations that
824 do not originate from file-backed memory mappings. Typical
825 applications that generate code, for example JITs in web browsers,
826 generate code into anonymous mmaped areas, whereas the "fixed" code
827 of the browser always lives in file-backed mappings.
828 --smc-check=all-non-file takes advantage of this observation,
829 limiting the overhead of checking to code which is likely to be JIT
830 generated.
831 --read-inline-info=<yes|no> [default: see below]
833 When enabled, Valgrind will read information about inlined function
834 calls from DWARF3 debug info. This slows Valgrind startup and makes
835 it use more memory (typically for each inlined piece of code, 6
836 words and space for the function name), but it results in more
837 descriptive stacktraces. Currently, this functionality is enabled
838 by default only for Linux, Android and Solaris targets and only for
839 the tools Memcheck, Massif, Helgrind and DRD. Here is an example of
840 some stacktraces with --read-inline-info=no:
841 ==15380== Conditional jump or move depends on uninitialised value(s)
843 ==15380== at 0x80484EA: main (inlinfo.c:6)
844 ==15380==
845 ==15380== Conditional jump or move depends on uninitialised value(s)
846 ==15380== at 0x8048550: fun_noninline (inlinfo.c:6)
847 ==15380== by 0x804850E: main (inlinfo.c:34)
848 ==15380==
849 ==15380== Conditional jump or move depends on uninitialised value(s)
850 ==15380== at 0x8048520: main (inlinfo.c:6)
851 And here are the same errors with --read-inline-info=yes:
853 ==15377== Conditional jump or move depends on uninitialised value(s)
855 ==15377== at 0x80484EA: fun_d (inlinfo.c:6)
856 ==15377== by 0x80484EA: fun_c (inlinfo.c:14)
857 ==15377== by 0x80484EA: fun_b (inlinfo.c:20)
858 ==15377== by 0x80484EA: fun_a (inlinfo.c:26)
859 ==15377== by 0x80484EA: main (inlinfo.c:33)
860 ==15377==
861 ==15377== Conditional jump or move depends on uninitialised value(s)
862 ==15377== at 0x8048550: fun_d (inlinfo.c:6)
863 ==15377== by 0x8048550: fun_noninline (inlinfo.c:41)
864 ==15377== by 0x804850E: main (inlinfo.c:34)
865 ==15377==
866 ==15377== Conditional jump or move depends on uninitialised value(s)
867 ==15377== at 0x8048520: fun_d (inlinfo.c:6)
868 ==15377== by 0x8048520: main (inlinfo.c:35)
869 --read-var-info=<yes|no> [default: no]
871 When enabled, Valgrind will read information about variable types
872 and locations from DWARF3 debug info. This slows Valgrind startup
873 significantly and makes it use significantly more memory, but for
874 the tools that can take advantage of it (Memcheck, Helgrind, DRD)
875 it can result in more precise error messages. For example, here are
876 some standard errors issued by Memcheck:
877 ==15363== Uninitialised byte(s) found during client check request
879 ==15363== at 0x80484A9: croak (varinfo1.c:28)
880 ==15363== by 0x8048544: main (varinfo1.c:55)
881 ==15363== Address 0x80497f7 is 7 bytes inside data symbol "global_i2"
882 ==15363==
883 ==15363== Uninitialised byte(s) found during client check request
884 ==15363== at 0x80484A9: croak (varinfo1.c:28)
885 ==15363== by 0x8048550: main (varinfo1.c:56)
886 ==15363== Address 0xbea0d0cc is on thread 1's stack
887 ==15363== in frame #1, created by main (varinfo1.c:45)
888 And here are the same errors with --read-var-info=yes:
890 ==15370== Uninitialised byte(s) found during client check request
892 ==15370== at 0x80484A9: croak (varinfo1.c:28)
893 ==15370== by 0x8048544: main (varinfo1.c:55)
894 ==15370== Location 0x80497f7 is 0 bytes inside global_i2[7],
895 ==15370== a global variable declared at varinfo1.c:41
896 ==15370==
897 ==15370== Uninitialised byte(s) found during client check request
898 ==15370== at 0x80484A9: croak (varinfo1.c:28)
899 ==15370== by 0x8048550: main (varinfo1.c:56)
900 ==15370== Location 0xbeb4a0cc is 0 bytes inside local var "local"
901 ==15370== declared at varinfo1.c:46, in frame #1 of thread 1
902 --vgdb-poll=<number> [default: 5000]
904 As part of its main loop, the Valgrind scheduler will poll to check
905 if some activity (such as an external command or some input from a
906 gdb) has to be handled by gdbserver. This activity poll will be
907 done after having run the given number of basic blocks (or slightly
908 more than the given number of basic blocks). This poll is quite
909 cheap so the default value is set relatively low. You might further
910 decrease this value if vgdb cannot use ptrace system call to
911 interrupt Valgrind if all threads are (most of the time) blocked in
912 a system call.
913 --vgdb-shadow-registers=no|yes [default: no]
915 When activated, gdbserver will expose the Valgrind shadow registers
916 to GDB. With this, the value of the Valgrind shadow registers can
917 be examined or changed using GDB. Exposing shadow registers only
918 works with GDB version 7.1 or later.
919 --vgdb-prefix=<prefix> [default: /tmp/vgdb-pipe]
921 To communicate with gdb/vgdb, the Valgrind gdbserver creates 3
922 files (2 named FIFOs and a mmap shared memory file). The prefix
923 option controls the directory and prefix for the creation of these
924 files.
925 --run-libc-freeres=<yes|no> [default: yes]
927 This option is only relevant when running Valgrind on Linux.
928 The GNU C library (libc.so), which is used by all programs, may
930 allocate memory for its own uses. Usually it doesn't bother to free
931 that memory when the program ends—there would be no point, since
932 the Linux kernel reclaims all process resources when a process
933 exits anyway, so it would just slow things down.
934 The glibc authors realised that this behaviour causes leak
936 checkers, such as Valgrind, to falsely report leaks in glibc, when
937 a leak check is done at exit. In order to avoid this, they provided
938 a routine called __libc_freeres specifically to make glibc release
939 all memory it has allocated. Memcheck therefore tries to run
940 __libc_freeres at exit.
941 Unfortunately, in some very old versions of glibc, __libc_freeres
943 is sufficiently buggy to cause segmentation faults. This was
944 particularly noticeable on Red Hat 7.1. So this option is provided
945 in order to inhibit the run of __libc_freeres. If your program
946 seems to run fine on Valgrind, but segfaults at exit, you may find
947 that --run-libc-freeres=no fixes that, although at the cost of
948 possibly falsely reporting space leaks in libc.so.
949 --run-cxx-freeres=<yes|no> [default: yes]
951 This option is only relevant when running Valgrind on Linux or
952 Solaris C++ programs.
953 The GNU Standard C++ library (libstdc++.so), which is used by all
955 C++ programs compiled with g++, may allocate memory for its own
956 uses. Usually it doesn't bother to free that memory when the
957 program ends—there would be no point, since the kernel reclaims all
958 process resources when a process exits anyway, so it would just
959 slow things down.
960 The gcc authors realised that this behaviour causes leak checkers,
962 such as Valgrind, to falsely report leaks in libstdc++, when a leak
963 check is done at exit. In order to avoid this, they provided a
964 routine called __gnu_cxx::__freeres specifically to make libstdc++
965 release all memory it has allocated. Memcheck therefore tries to
966 run __gnu_cxx::__freeres at exit.
967 For the sake of flexibility and unforeseen problems with
969 __gnu_cxx::__freeres, option --run-cxx-freeres=no exists, although
970 at the cost of possibly falsely reporting space leaks in
971 libstdc++.so.
972 --sim-hints=hint1,hint2,...
974 Pass miscellaneous hints to Valgrind which slightly modify the
975 simulated behaviour in nonstandard or dangerous ways, possibly to
976 help the simulation of strange features. By default no hints are
977 enabled. Use with caution! Currently known hints are:
978 · lax-ioctls: Be very lax about ioctl handling; the only
980 assumption is that the size is correct. Doesn't require the
981 full buffer to be initialised when writing. Without this, using
982 some device drivers with a large number of strange ioctl
983 commands becomes very tiresome.
984 · fuse-compatible: Enable special handling for certain system
986 calls that may block in a FUSE file-system. This may be
987 necessary when running Valgrind on a multi-threaded program
988 that uses one thread to manage a FUSE file-system and another
989 thread to access that file-system.
990 · enable-outer: Enable some special magic needed when the program
992 being run is itself Valgrind.
993 · no-inner-prefix: Disable printing a prefix > in front of each
995 stdout or stderr output line in an inner Valgrind being run by
996 an outer Valgrind. This is useful when running Valgrind
997 regression tests in an outer/inner setup. Note that the prefix
998 > will always be printed in front of the inner debug logging
999 lines.
1000 · no-nptl-pthread-stackcache: This hint is only relevant when
1002 running Valgrind on Linux; it is ignored on Solaris and Mac OS
1003 X.
1004 The GNU glibc pthread library (libpthread.so), which is used by
1006 pthread programs, maintains a cache of pthread stacks. When a
1007 pthread terminates, the memory used for the pthread stack and
1008 some thread local storage related data structure are not always
1009 directly released. This memory is kept in a cache (up to a
1010 certain size), and is re-used if a new thread is started.
1011 This cache causes the helgrind tool to report some false
1013 positive race condition errors on this cached memory, as
1014 helgrind does not understand the internal glibc cache
1015 synchronisation primitives. So, when using helgrind, disabling
1016 the cache helps to avoid false positive race conditions, in
1017 particular when using thread local storage variables (e.g.
1018 variables using the __thread qualifier).
1019 When using the memcheck tool, disabling the cache ensures the
1021 memory used by glibc to handle __thread variables is directly
1022 released when a thread terminates.
1023 Note: Valgrind disables the cache using some internal knowledge
1025 of the glibc stack cache implementation and by examining the
1026 debug information of the pthread library. This technique is
1027 thus somewhat fragile and might not work for all glibc
1028 versions. This has been successfully tested with various glibc
1029 versions (e.g. 2.11, 2.16, 2.18) on various platforms.
1030 · lax-doors: (Solaris only) Be very lax about door syscall
1032 handling over unrecognised door file descriptors. Does not
1033 require that full buffer is initialised when writing. Without
1034 this, programs using libdoor(3LIB) functionality with
1035 completely proprietary semantics may report large number of
1036 false positives.
1037 · fallback-llsc: (MIPS and ARM64 only): Enables an alternative
1039 implementation of Load-Linked (LL) and Store-Conditional (SC)
1040 instructions. The standard implementation gives more correct
1041 behaviour, but can cause indefinite looping on certain
1042 processor implementations that are intolerant of extra memory
1043 references between LL and SC. So far this is known only to
1044 happen on Cavium 3 cores. You should not need to use this flag,
1045 since the relevant cores are detected at startup and the
1046 alternative implementation is automatically enabled if
1047 necessary. There is no equivalent anti-flag: you cannot
1048 force-disable the alternative implementation, if it is
1049 automatically enabled. The underlying problem exists because
1050 the "standard" implementation of LL and SC is done by copying
1051 through LL and SC instructions into the instrumented code.
1052 However, tools may insert extra instrumentation memory
1053 references in between the LL and SC instructions. These memory
1054 references are not present in the original uninstrumented code,
1055 and their presence in the instrumented code can cause the SC
1056 instructions to persistently fail, leading to indefinite
1057 looping in LL-SC blocks. The alternative implementation gives
1058 correct behaviour of LL and SC instructions between threads in
1059 a process, up to and including the ABA scenario. It also gives
1060 correct behaviour between a Valgrinded thread and a
1061 non-Valgrinded thread running in a different process, that
1062 communicate via shared memory, but only up to and including
1063 correct CAS behaviour -- in this case the ABA scenario may not
1064 be correctly handled.
1065 --fair-sched=<no|yes|try> [default: no]
1067 The --fair-sched option controls the locking mechanism used by
1068 Valgrind to serialise thread execution. The locking mechanism
1069 controls the way the threads are scheduled, and different settings
1070 give different trade-offs between fairness and performance. For
1071 more details about the Valgrind thread serialisation scheme and its
1072 impact on performance and thread scheduling, see Scheduling and
1073 Multi-Thread Performance.
1074 · The value --fair-sched=yes activates a fair scheduler. In
1076 short, if multiple threads are ready to run, the threads will
1077 be scheduled in a round robin fashion. This mechanism is not
1078 available on all platforms or Linux versions. If not available,
1079 using --fair-sched=yes will cause Valgrind to terminate with an
1080 error.
1081 You may find this setting improves overall responsiveness if
1083 you are running an interactive multithreaded program, for
1084 example a web browser, on Valgrind.
1085 · The value --fair-sched=try activates fair scheduling if
1087 available on the platform. Otherwise, it will automatically
1088 fall back to --fair-sched=no.
1089 · The value --fair-sched=no activates a scheduler which does not
1091 guarantee fairness between threads ready to run, but which in
1092 general gives the highest performance.
1093 --kernel-variant=variant1,variant2,...
1095 Handle system calls and ioctls arising from minor variants of the
1096 default kernel for this platform. This is useful for running on
1097 hacked kernels or with kernel modules which support nonstandard
1098 ioctls, for example. Use with caution. If you don't understand what
1099 this option does then you almost certainly don't need it. Currently
1100 known variants are:
1101 · bproc: support the sys_broc system call on x86. This is for
1103 running on BProc, which is a minor variant of standard Linux
1104 which is sometimes used for building clusters.
1105 · android-no-hw-tls: some versions of the Android emulator for
1107 ARM do not provide a hardware TLS (thread-local state)
1108 register, and Valgrind crashes at startup. Use this variant to
1109 select software support for TLS.
1110 · android-gpu-sgx5xx: use this to support handling of proprietary
1112 ioctls for the PowerVR SGX 5XX series of GPUs on Android
1113 devices. Failure to select this does not cause stability
1114 problems, but may cause Memcheck to report false errors after
1115 the program performs GPU-specific ioctls.
1116 · android-gpu-adreno3xx: similarly, use this to support handling
1118 of proprietary ioctls for the Qualcomm Adreno 3XX series of
1119 GPUs on Android devices.
1120 --merge-recursive-frames=<number> [default: 0]
1122 Some recursive algorithms, for example balanced binary tree
1123 implementations, create many different stack traces, each
1124 containing cycles of calls. A cycle is defined as two identical
1125 program counter values separated by zero or more other program
1126 counter values. Valgrind may then use a lot of memory to store all
1127 these stack traces. This is a poor use of memory considering that
1128 such stack traces contain repeated uninteresting recursive calls
1129 instead of more interesting information such as the function that
1130 has initiated the recursive call.
1131 The option --merge-recursive-frames=<number> instructs Valgrind to
1133 detect and merge recursive call cycles having a size of up to
1134 <number> frames. When such a cycle is detected, Valgrind records
1135 the cycle in the stack trace as a unique program counter.
1136 The value 0 (the default) causes no recursive call merging. A value
1138 of 1 will cause stack traces of simple recursive algorithms (for
1139 example, a factorial implementation) to be collapsed. A value of 2
1140 will usually be needed to collapse stack traces produced by
1141 recursive algorithms such as binary trees, quick sort, etc. Higher
1142 values might be needed for more complex recursive algorithms.
1143 Note: recursive calls are detected by analysis of program counter
1145 values. They are not detected by looking at function names.
1146 --num-transtab-sectors=<number> [default: 6 for Android platforms, 16
1148 for all others]
1149 Valgrind translates and instruments your program's machine code in
1150 small fragments (basic blocks). The translations are stored in a
1151 translation cache that is divided into a number of sections
1152 (sectors). If the cache is full, the sector containing the oldest
1153 translations is emptied and reused. If these old translations are
1154 needed again, Valgrind must re-translate and re-instrument the
1155 corresponding machine code, which is expensive. If the "executed
1156 instructions" working set of a program is big, increasing the
1157 number of sectors may improve performance by reducing the number of
1158 re-translations needed. Sectors are allocated on demand. Once
1159 allocated, a sector can never be freed, and occupies considerable
1160 space, depending on the tool and the value of
1161 --avg-transtab-entry-size (about 40 MB per sector for Memcheck).
1162 Use the option --stats=yes to obtain precise information about the
1163 memory used by a sector and the allocation and recycling of
1164 sectors.
1165 --avg-transtab-entry-size=<number> [default: 0, meaning use tool
1167 provided default]
1168 Average size of translated basic block. This average size is used
1169 to dimension the size of a sector. Each tool provides a default
1170 value to be used. If this default value is too small, the
1171 translation sectors will become full too quickly. If this default
1172 value is too big, a significant part of the translation sector
1173 memory will be unused. Note that the average size of a basic block
1174 translation depends on the tool, and might depend on tool options.
1175 For example, the memcheck option --track-origins=yes increases the
1176 size of the basic block translations. Use --avg-transtab-entry-size
1177 to tune the size of the sectors, either to gain memory or to avoid
1178 too many retranslations.
1179 --aspace-minaddr=<address> [default: depends on the platform]
1181 To avoid potential conflicts with some system libraries, Valgrind
1182 does not use the address space below --aspace-minaddr value,
1183 keeping it reserved in case a library specifically requests memory
1184 in this region. So, some "pessimistic" value is guessed by Valgrind
1185 depending on the platform. On linux, by default, Valgrind avoids
1186 using the first 64MB even if typically there is no conflict in this
1187 complete zone. You can use the option --aspace-minaddr to have your
1188 memory hungry application benefitting from more of this lower
1189 memory. On the other hand, if you encounter a conflict, increasing
1190 aspace-minaddr value might solve it. Conflicts will typically
1191 manifest themselves with mmap failures in the low range of the
1192 address space. The provided address must be page aligned and must
1193 be equal or bigger to 0x1000 (4KB). To find the default value on
1194 your platform, do something such as valgrind -d -d date 2>&1 | grep
1195 -i minaddr. Values lower than 0x10000 (64KB) are known to create
1196 problems on some distributions.
1197 --valgrind-stacksize=<number> [default: 1MB]
1199 For each thread, Valgrind needs its own 'private' stack. The
1200 default size for these stacks is largely dimensioned, and so should
1201 be sufficient in most cases. In case the size is too small,
1202 Valgrind will segfault. Before segfaulting, a warning might be
1203 produced by Valgrind when approaching the limit.
1204 Use the option --valgrind-stacksize if such an (unlikely) warning
1206 is produced, or Valgrind dies due to a segmentation violation. Such
1207 segmentation violations have been seen when demangling huge C++
1208 symbols.
1209 If your application uses many threads and needs a lot of memory,
1211 you can gain some memory by reducing the size of these Valgrind
1212 stacks using the option --valgrind-stacksize.
1213 --show-emwarns=<yes|no> [default: no]
1215 When enabled, Valgrind will emit warnings about its CPU emulation
1216 in certain cases. These are usually not interesting.
1217 --require-text-symbol=:sonamepatt:fnnamepatt
1219 When a shared object whose soname matches sonamepatt is loaded into
1220 the process, examine all the text symbols it exports. If none of
1221 those match fnnamepatt, print an error message and abandon the run.
1222 This makes it possible to ensure that the run does not continue
1223 unless a given shared object contains a particular function name.
1224 Both sonamepatt and fnnamepatt can be written using the usual ?
1226 and * wildcards. For example: ":*libc.so*:foo?bar". You may use
1227 characters other than a colon to separate the two patterns. It is
1228 only important that the first character and the separator character
1229 are the same. For example, the above example could also be written
1230 "Q*libc.so*Qfoo?bar". Multiple
1231 --require-text-symbol flags are allowed, in which case shared
1232 objects that are loaded into the process will be checked against
1233 all of them.
1234 The purpose of this is to support reliable usage of marked-up
1236 libraries. For example, suppose we have a version of GCC's
1237 libgomp.so which has been marked up with annotations to support
1238 Helgrind. It is only too easy and confusing to load the wrong,
1239 un-annotated libgomp.so into the application. So the idea is: add a
1240 text symbol in the marked-up library, for example
1241 annotated_for_helgrind_3_6, and then give the flag
1242 --require-text-symbol=:*libgomp*so*:annotated_for_helgrind_3_6 so
1243 that when libgomp.so is loaded, Valgrind scans its symbol table,
1244 and if the symbol isn't present the run is aborted, rather than
1245 continuing silently with the un-marked-up library. Note that you
1246 should put the entire flag in quotes to stop shells expanding up
1247 the * and ? wildcards.
1248 --soname-synonyms=syn1=pattern1,syn2=pattern2,...
1250 When a shared library is loaded, Valgrind checks for functions in
1251 the library that must be replaced or wrapped. For example, Memcheck
1252 replaces some string and memory functions (strchr, strlen, strcpy,
1253 memchr, memcpy, memmove, etc.) with its own versions. Such
1254 replacements are normally done only in shared libraries whose
1255 soname matches a predefined soname pattern (e.g. libc.so* on
1256 linux). By default, no replacement is done for a statically linked
1257 binary or for alternative libraries, except for the allocation
1258 functions (malloc, free, calloc, memalign, realloc, operator new,
1259 operator delete, etc.) Such allocation functions are intercepted by
1260 default in any shared library or in the executable if they are
1261 exported as global symbols. This means that if a replacement
1262 allocation library such as tcmalloc is found, its functions are
1263 also intercepted by default. In some cases, the replacements allow
1264 --soname-synonyms to specify one additional synonym pattern, giving
1265 flexibility in the replacement. Or to prevent interception of all
1266 public allocation symbols.
1267 Currently, this flexibility is only allowed for the malloc related
1269 functions, using the synonym somalloc. This synonym is usable for
1270 all tools doing standard replacement of malloc related functions
1271 (e.g. memcheck, helgrind, drd, massif, dhat, exp-sgcheck).
1272 · Alternate malloc library: to replace the malloc related
1274 functions in a specific alternate library with soname
1275 mymalloclib.so (and not in any others), give the option
1276 --soname-synonyms=somalloc=mymalloclib.so. A pattern can be
1277 used to match multiple libraries sonames. For example,
1278 --soname-synonyms=somalloc=*tcmalloc* will match the soname of
1279 all variants of the tcmalloc library (native, debug, profiled,
1280 ... tcmalloc variants).
1281 Note: the soname of a elf shared library can be retrieved using
1283 the readelf utility.
1284 · Replacements in a statically linked library are done by using
1286 the NONE pattern. For example, if you link with libtcmalloc.a,
1287 and only want to intercept the malloc related functions in the
1288 executable (and standard libraries) themselves, but not any
1289 other shared libraries, you can give the option
1290 --soname-synonyms=somalloc=NONE. Note that a NONE pattern will
1291 match the main executable and any shared library having no
1292 soname.
1293 · To run a "default" Firefox build for Linux, in which JEMalloc
1295 is linked in to the main executable, use
1296 --soname-synonyms=somalloc=NONE.
1297 · To only intercept allocation symbols in the default system
1299 libraries, but not in any other shared library or the
1300 executable defining public malloc or operator new related
1301 functions use a non-existing library name like
1302 --soname-synonyms=somalloc=nouserintercepts (where
1303 nouserintercepts can be any non-existing library name).
1304 · Shared library of the dynamic (runtime) linker is excluded from
1306 searching for global public symbols, such as those for the
1307 malloc related functions (identified by somalloc synonym).
1308 --progress-interval=<number> [default: 0, meaning 'disabled']
1310 This is an enhancement to Valgrind's debugging output. It is
1311 unlikely to be of interest to end users.
1312 When number is set to a non-zero value, Valgrind will print a
1314 one-line progress summary every number seconds. Valid settings for
1315 number are between 0 and 3600 inclusive. Here's some example output
1316 with number set to 10:
1317 PROGRESS: U 110s, W 113s, 97.3% CPU, EvC 414.79M, TIn 616.7k, TOut 0.5k, #thr 67
1319 PROGRESS: U 120s, W 124s, 96.8% CPU, EvC 505.27M, TIn 636.6k, TOut 3.0k, #thr 64
1320 PROGRESS: U 130s, W 134s, 97.0% CPU, EvC 574.90M, TIn 657.5k, TOut 3.0k, #thr 63
1321 Each line shows:
1323 · U: total user time.RE
1325 · W: total wallclock time.RE
1327 · CPU: overall average cpu use.RE
1329 · EvC: number of event checks. An event
1331 check is a backwards branch in the
1332 simulated program, so this is a
1333 measure of forward progress of the
1334 program.RE
1335 · TIn: number of code blocks instrumented
1337 by the JIT.RE
1338 · TOut: number of instrumented code
1340 blocks that have been thrown
1341 away.RE
1342 · #thr: number of threads in the
1344 program.RE
1345 From the progress of these, it is
1347 possible to observe:
1348 · when the program is compute bound
1350 (TIn
1351 rises slowly, EvC rises
1352 rapidly).RE
1353 · when the program is in a
1355 spinloop
1356 (TIn/TOut
1357 fixed, EvC rises
1358 rapidly).RE
1359 · when the program is
1361 JIT-bound (TIn
1362 rises rapidly).RE
1363 · when the program is
1365 rapidly discarding code
1366 (TOut rises
1367 rapidly).RE
1368 · when the program is
1370 about to achieve
1371 some expected state
1372 (EvC
1373 arrives at some
1374 value you
1375 expect).RE
1376 · when the
1378 program is
1379 idling (U rises
1380 more
1381 slowly than
1382 W).RE
1383
1386 There are also some options for debugging Valgrind itself. You
1387 shouldn't need to use them in the normal run of things. If you wish to
1388 see the list, use the --help-debug option.
1389
1391 --leak-check=<no|summary|yes|full> [default: summary]
1392 When enabled, search for memory leaks when the client program
1393 finishes. If set to summary, it says how many leaks occurred. If
1394 set to full or yes, each individual leak will be shown in detail
1395 and/or counted as an error, as specified by the options
1396 --show-leak-kinds and --errors-for-leak-kinds.
1397 --leak-resolution=<low|med|high> [default: high]
1399 When doing leak checking, determines how willing Memcheck is to
1400 consider different backtraces to be the same for the purposes of
1401 merging multiple leaks into a single leak report. When set to low,
1402 only the first two entries need match. When med, four entries have
1403 to match. When high, all entries need to match.
1404 For hardcore leak debugging, you probably want to use
1406 --leak-resolution=high together with --num-callers=40 or some such
1407 large number.
1408 Note that the --leak-resolution setting does not affect Memcheck's
1410 ability to find leaks. It only changes how the results are
1411 presented.
1412 --show-leak-kinds=<set> [default: definite,possible]
1414 Specifies the leak kinds to show in a full leak search, in one of
1415 the following ways:
1416 · a comma separated list of one or more of definite indirect
1418 possible reachable.
1419 · all to specify the complete set (all leak kinds). It is
1421 equivalent to
1422 --show-leak-kinds=definite,indirect,possible,reachable.
1423 · none for the empty set.
1425 --errors-for-leak-kinds=<set> [default: definite,possible]
1427 Specifies the leak kinds to count as errors in a full leak search.
1428 The <set> is specified similarly to --show-leak-kinds
1429 --leak-check-heuristics=<set> [default: all]
1431 Specifies the set of leak check heuristics to be used during leak
1432 searches. The heuristics control which interior pointers to a block
1433 cause it to be considered as reachable. The heuristic set is
1434 specified in one of the following ways:
1435 · a comma separated list of one or more of stdstring length64
1437 newarray multipleinheritance.
1438 · all to activate the complete set of heuristics. It is
1440 equivalent to
1441 --leak-check-heuristics=stdstring,length64,newarray,multipleinheritance.
1442 · none for the empty set.
1444 Note that these heuristics are dependent on the layout of the
1446 objects produced by the C++ compiler. They have been tested with
1447 some gcc versions (e.g. 4.4 and 4.7). They might not work properly
1448 with other C++ compilers.
1449 --show-reachable=<yes|no> , --show-possibly-lost=<yes|no>
1451 These options provide an alternative way to specify the leak kinds
1452 to show:
1453 · --show-reachable=no --show-possibly-lost=yes is equivalent to
1455 --show-leak-kinds=definite,possible.
1456 · --show-reachable=no --show-possibly-lost=no is equivalent to
1458 --show-leak-kinds=definite.
1459 · --show-reachable=yes is equivalent to --show-leak-kinds=all.
1461 Note that --show-possibly-lost=no has no effect if
1463 --show-reachable=yes is specified.
1464 --xtree-leak=<no|yes> [no]
1466 If set to yes, the results for the leak search done at exit will be
1467 output in a 'Callgrind Format' execution tree file. Note that this
1468 automatically sets the options --leak-check=full and
1469 --show-leak-kinds=all, to allow xtree visualisation tools such as
1470 kcachegrind to select what kind to leak to visualise. The produced
1471 file will contain the following events:
1472 · RB : Reachable Bytes
1474 · PB : Possibly lost Bytes
1476 · IB : Indirectly lost Bytes
1478 · DB : Definitely lost Bytes (direct plus indirect)
1480 · DIB : Definitely Indirectly lost Bytes (subset of DB)
1482 · RBk : reachable Blocks
1484 · PBk : Possibly lost Blocks
1486 · IBk : Indirectly lost Blocks
1488 · DBk : Definitely lost Blocks
1490 The increase or decrease for all events above will also be output
1492 in the file to provide the delta (increase or decrease) between 2
1493 successive leak searches. For example, iRB is the increase of the
1494 RB event, dPBk is the decrease of PBk event. The values for the
1495 increase and decrease events will be zero for the first leak search
1496 done.
1497 See ??? for a detailed explanation about execution trees.
1499 --xtree-leak-file=<filename> [default: xtleak.kcg.%p]
1501 Specifies that Valgrind should produce the xtree leak report in the
1502 specified file. Any %p, %q or %n sequences appearing in the
1503 filename are expanded in exactly the same way as they are for
1504 --log-file. See the description of --log-file for details.
1505 See ??? for a detailed explanation about execution trees formats.
1507 --undef-value-errors=<yes|no> [default: yes]
1509 Controls whether Memcheck reports uses of undefined value errors.
1510 Set this to no if you don't want to see undefined value errors. It
1511 also has the side effect of speeding up Memcheck somewhat.
1512 AddrCheck (removed in Valgrind 3.1.0) functioned like Memcheck with
1513 --undef-value-errors=no.
1514 --track-origins=<yes|no> [default: no]
1516 Controls whether Memcheck tracks the origin of uninitialised
1517 values. By default, it does not, which means that although it can
1518 tell you that an uninitialised value is being used in a dangerous
1519 way, it cannot tell you where the uninitialised value came from.
1520 This often makes it difficult to track down the root problem.
1521 When set to yes, Memcheck keeps track of the origins of all
1523 uninitialised values. Then, when an uninitialised value error is
1524 reported, Memcheck will try to show the origin of the value. An
1525 origin can be one of the following four places: a heap block, a
1526 stack allocation, a client request, or miscellaneous other sources
1527 (eg, a call to brk).
1528 For uninitialised values originating from a heap block, Memcheck
1530 shows where the block was allocated. For uninitialised values
1531 originating from a stack allocation, Memcheck can tell you which
1532 function allocated the value, but no more than that -- typically it
1533 shows you the source location of the opening brace of the function.
1534 So you should carefully check that all of the function's local
1535 variables are initialised properly.
1536 Performance overhead: origin tracking is expensive. It halves
1538 Memcheck's speed and increases memory use by a minimum of 100MB,
1539 and possibly more. Nevertheless it can drastically reduce the
1540 effort required to identify the root cause of uninitialised value
1541 errors, and so is often a programmer productivity win, despite
1542 running more slowly.
1543 Accuracy: Memcheck tracks origins quite accurately. To avoid very
1545 large space and time overheads, some approximations are made. It is
1546 possible, although unlikely, that Memcheck will report an incorrect
1547 origin, or not be able to identify any origin.
1548 Note that the combination --track-origins=yes and
1550 --undef-value-errors=no is nonsensical. Memcheck checks for and
1551 rejects this combination at startup.
1552 --partial-loads-ok=<yes|no> [default: yes]
1554 Controls how Memcheck handles 32-, 64-, 128- and 256-bit naturally
1555 aligned loads from addresses for which some bytes are addressable
1556 and others are not. When yes, such loads do not produce an address
1557 error. Instead, loaded bytes originating from illegal addresses are
1558 marked as uninitialised, and those corresponding to legal addresses
1559 are handled in the normal way.
1560 When no, loads from partially invalid addresses are treated the
1562 same as loads from completely invalid addresses: an illegal-address
1563 error is issued, and the resulting bytes are marked as initialised.
1564 Note that code that behaves in this way is in violation of the ISO
1566 C/C++ standards, and should be considered broken. If at all
1567 possible, such code should be fixed.
1568 --expensive-definedness-checks=<no|auto|yes> [default: auto]
1570 Controls whether Memcheck should employ more precise but also more
1571 expensive (time consuming) instrumentation when checking the
1572 definedness of certain values. In particular, this affects the
1573 instrumentation of integer adds, subtracts and equality
1574 comparisons.
1575 Selecting --expensive-definedness-checks=yes causes Memcheck to use
1577 the most accurate analysis possible. This minimises false error
1578 rates but can cause up to 30% performance degradation.
1579 Selecting --expensive-definedness-checks=no causes Memcheck to use
1581 the cheapest instrumentation possible. This maximises performance
1582 but will normally give an unusably high false error rate.
1583 The default setting, --expensive-definedness-checks=auto, is
1585 strongly recommended. This causes Memcheck to use the minimum of
1586 expensive instrumentation needed to achieve the same false error
1587 rate as --expensive-definedness-checks=yes. It also enables an
1588 instrumentation-time analysis pass which aims to further reduce the
1589 costs of accurate instrumentation. Overall, the performance loss is
1590 generally around 5% relative to --expensive-definedness-checks=no,
1591 although this is strongly workload dependent. Note that the exact
1592 instrumentation settings in this mode are architecture dependent.
1593 --keep-stacktraces=alloc|free|alloc-and-free|alloc-then-free|none
1595 [default: alloc-and-free]
1596 Controls which stack trace(s) to keep for malloc'd and/or free'd
1597 blocks.
1598 With alloc-then-free, a stack trace is recorded at allocation time,
1600 and is associated with the block. When the block is freed, a second
1601 stack trace is recorded, and this replaces the allocation stack
1602 trace. As a result, any "use after free" errors relating to this
1603 block can only show a stack trace for where the block was freed.
1604 With alloc-and-free, both allocation and the deallocation stack
1606 traces for the block are stored. Hence a "use after free" error
1607 will show both, which may make the error easier to diagnose.
1608 Compared to alloc-then-free, this setting slightly increases
1609 Valgrind's memory use as the block contains two references instead
1610 of one.
1611 With alloc, only the allocation stack trace is recorded (and
1613 reported). With free, only the deallocation stack trace is recorded
1614 (and reported). These values somewhat decrease Valgrind's memory
1615 and cpu usage. They can be useful depending on the error types you
1616 are searching for and the level of detail you need to analyse them.
1617 For example, if you are only interested in memory leak errors, it
1618 is sufficient to record the allocation stack traces.
1619 With none, no stack traces are recorded for malloc and free
1621 operations. If your program allocates a lot of blocks and/or
1622 allocates/frees from many different stack traces, this can
1623 significantly decrease cpu and/or memory required. Of course, few
1624 details will be reported for errors related to heap blocks.
1625 Note that once a stack trace is recorded, Valgrind keeps the stack
1627 trace in memory even if it is not referenced by any block. Some
1628 programs (for example, recursive algorithms) can generate a huge
1629 number of stack traces. If Valgrind uses too much memory in such
1630 circumstances, you can reduce the memory required with the options
1631 --keep-stacktraces and/or by using a smaller value for the option
1632 --num-callers.
1633 If you want to use --xtree-memory=full memory profiling (see ???
1635 ), then you cannot specify --keep-stacktraces=free or
1636 --keep-stacktraces=none.
1637 --freelist-vol=<number> [default: 20000000]
1639 When the client program releases memory using free (in C) or delete
1640 (C++), that memory is not immediately made available for
1641 re-allocation. Instead, it is marked inaccessible and placed in a
1642 queue of freed blocks. The purpose is to defer as long as possible
1643 the point at which freed-up memory comes back into circulation.
1644 This increases the chance that Memcheck will be able to detect
1645 invalid accesses to blocks for some significant period of time
1646 after they have been freed.
1647 This option specifies the maximum total size, in bytes, of the
1649 blocks in the queue. The default value is twenty million bytes.
1650 Increasing this increases the total amount of memory used by
1651 Memcheck but may detect invalid uses of freed blocks which would
1652 otherwise go undetected.
1653 --freelist-big-blocks=<number> [default: 1000000]
1655 When making blocks from the queue of freed blocks available for
1656 re-allocation, Memcheck will in priority re-circulate the blocks
1657 with a size greater or equal to --freelist-big-blocks. This ensures
1658 that freeing big blocks (in particular freeing blocks bigger than
1659 --freelist-vol) does not immediately lead to a re-circulation of
1660 all (or a lot of) the small blocks in the free list. In other
1661 words, this option increases the likelihood to discover dangling
1662 pointers for the "small" blocks, even when big blocks are freed.
1663 Setting a value of 0 means that all the blocks are re-circulated in
1665 a FIFO order.
1666 --workaround-gcc296-bugs=<yes|no> [default: no]
1668 When enabled, assume that reads and writes some small distance
1669 below the stack pointer are due to bugs in GCC 2.96, and does not
1670 report them. The "small distance" is 256 bytes by default. Note
1671 that GCC 2.96 is the default compiler on some ancient Linux
1672 distributions (RedHat 7.X) and so you may need to use this option.
1673 Do not use it if you do not have to, as it can cause real errors to
1674 be overlooked. A better alternative is to use a more recent GCC in
1675 which this bug is fixed.
1676 You may also need to use this option when working with GCC 3.X or
1678 4.X on 32-bit PowerPC Linux. This is because GCC generates code
1679 which occasionally accesses below the stack pointer, particularly
1680 for floating-point to/from integer conversions. This is in
1681 violation of the 32-bit PowerPC ELF specification, which makes no
1682 provision for locations below the stack pointer to be accessible.
1683 This option is deprecated as of version 3.12 and may be removed
1685 from future versions. You should instead use
1686 --ignore-range-below-sp to specify the exact range of offsets below
1687 the stack pointer that should be ignored. A suitable equivalent is
1688 --ignore-range-below-sp=1024-1.
1689 --ignore-range-below-sp=<number>-<number>
1691 This is a more general replacement for the deprecated
1692 --workaround-gcc296-bugs option. When specified, it causes Memcheck
1693 not to report errors for accesses at the specified offsets below
1694 the stack pointer. The two offsets must be positive decimal numbers
1695 and -- somewhat counterintuitively -- the first one must be larger,
1696 in order to imply a non-wraparound address range to ignore. For
1697 example, to ignore 4 byte accesses at 8192 bytes below the stack
1698 pointer, use --ignore-range-below-sp=8192-8189. Only one range may
1699 be specified.
1700 --show-mismatched-frees=<yes|no> [default: yes]
1702 When enabled, Memcheck checks that heap blocks are deallocated
1703 using a function that matches the allocating function. That is, it
1704 expects free to be used to deallocate blocks allocated by malloc,
1705 delete for blocks allocated by new, and delete[] for blocks
1706 allocated by new[]. If a mismatch is detected, an error is
1707 reported. This is in general important because in some
1708 environments, freeing with a non-matching function can cause
1709 crashes.
1710 There is however a scenario where such mismatches cannot be
1712 avoided. That is when the user provides implementations of
1713 new/new[] that call malloc and of delete/delete[] that call free,
1714 and these functions are asymmetrically inlined. For example,
1715 imagine that delete[] is inlined but new[] is not. The result is
1716 that Memcheck "sees" all delete[] calls as direct calls to free,
1717 even when the program source contains no mismatched calls.
1718 This causes a lot of confusing and irrelevant error reports.
1720 --show-mismatched-frees=no disables these checks. It is not
1721 generally advisable to disable them, though, because you may miss
1722 real errors as a result.
1723 --ignore-ranges=0xPP-0xQQ[,0xRR-0xSS]
1725 Any ranges listed in this option (and multiple ranges can be
1726 specified, separated by commas) will be ignored by Memcheck's
1727 addressability checking.
1728 --malloc-fill=<hexnumber>
1730 Fills blocks allocated by malloc, new, etc, but not by calloc, with
1731 the specified byte. This can be useful when trying to shake out
1732 obscure memory corruption problems. The allocated area is still
1733 regarded by Memcheck as undefined -- this option only affects its
1734 contents. Note that --malloc-fill does not affect a block of memory
1735 when it is used as argument to client requests
1736 VALGRIND_MEMPOOL_ALLOC or VALGRIND_MALLOCLIKE_BLOCK.
1737 --free-fill=<hexnumber>
1739 Fills blocks freed by free, delete, etc, with the specified byte
1740 value. This can be useful when trying to shake out obscure memory
1741 corruption problems. The freed area is still regarded by Memcheck
1742 as not valid for access -- this option only affects its contents.
1743 Note that --free-fill does not affect a block of memory when it is
1744 used as argument to client requests VALGRIND_MEMPOOL_FREE or
1745 VALGRIND_FREELIKE_BLOCK.
1746
1748 --I1=<size>,<associativity>,<line size>
1749 Specify the size, associativity and line size of the level 1
1750 instruction cache.
1751 --D1=<size>,<associativity>,<line size>
1753 Specify the size, associativity and line size of the level 1 data
1754 cache.
1755 --LL=<size>,<associativity>,<line size>
1757 Specify the size, associativity and line size of the last-level
1758 cache.
1759 --cache-sim=no|yes [yes]
1761 Enables or disables collection of cache access and miss counts.
1762 --branch-sim=no|yes [no]
1764 Enables or disables collection of branch instruction and
1765 misprediction counts. By default this is disabled as it slows
1766 Cachegrind down by approximately 25%. Note that you cannot specify
1767 --cache-sim=no and --branch-sim=no together, as that would leave
1768 Cachegrind with no information to collect.
1769 --cachegrind-out-file=<file>
1771 Write the profile data to file rather than to the default output
1772 file, cachegrind.out.<pid>. The %p and %q format specifiers can be
1773 used to embed the process ID and/or the contents of an environment
1774 variable in the name, as is the case for the core option --log-
1775 file.
1776
1778 --callgrind-out-file=<file>
1779 Write the profile data to file rather than to the default output
1780 file, callgrind.out.<pid>. The %p and %q format specifiers can be
1781 used to embed the process ID and/or the contents of an environment
1782 variable in the name, as is the case for the core option --log-
1783 file. When multiple dumps are made, the file name is modified
1784 further; see below.
1785 --dump-line=<no|yes> [default: yes]
1787 This specifies that event counting should be performed at source
1788 line granularity. This allows source annotation for sources which
1789 are compiled with debug information (-g).
1790 --dump-instr=<no|yes> [default: no]
1792 This specifies that event counting should be performed at
1793 per-instruction granularity. This allows for assembly code
1794 annotation. Currently the results can only be displayed by
1795 KCachegrind.
1796 --compress-strings=<no|yes> [default: yes]
1798 This option influences the output format of the profile data. It
1799 specifies whether strings (file and function names) should be
1800 identified by numbers. This shrinks the file, but makes it more
1801 difficult for humans to read (which is not recommended in any
1802 case).
1803 --compress-pos=<no|yes> [default: yes]
1805 This option influences the output format of the profile data. It
1806 specifies whether numerical positions are always specified as
1807 absolute values or are allowed to be relative to previous numbers.
1808 This shrinks the file size.
1809 --combine-dumps=<no|yes> [default: no]
1811 When enabled, when multiple profile data parts are to be generated
1812 these parts are appended to the same output file. Not recommended.
1813 --dump-every-bb=<count> [default: 0, never]
1815 Dump profile data every count basic blocks. Whether a dump is
1816 needed is only checked when Valgrind's internal scheduler is run.
1817 Therefore, the minimum setting useful is about 100000. The count is
1818 a 64-bit value to make long dump periods possible.
1819 --dump-before=<function>
1821 Dump when entering function.
1822 --zero-before=<function>
1824 Zero all costs when entering function.
1825 --dump-after=<function>
1827 Dump when leaving function.
1828 --instr-atstart=<yes|no> [default: yes]
1830 Specify if you want Callgrind to start simulation and profiling
1831 from the beginning of the program. When set to no, Callgrind will
1832 not be able to collect any information, including calls, but it
1833 will have at most a slowdown of around 4, which is the minimum
1834 Valgrind overhead. Instrumentation can be interactively enabled via
1835 callgrind_control -i on.
1836 Note that the resulting call graph will most probably not contain
1838 main, but will contain all the functions executed after
1839 instrumentation was enabled. Instrumentation can also
1840 programatically enabled/disabled. See the Callgrind include file
1841 callgrind.h for the macro you have to use in your source code.
1842 For cache simulation, results will be less accurate when switching
1844 on instrumentation later in the program run, as the simulator
1845 starts with an empty cache at that moment. Switch on event
1846 collection later to cope with this error.
1847 --collect-atstart=<yes|no> [default: yes]
1849 Specify whether event collection is enabled at beginning of the
1850 profile run.
1851 To only look at parts of your program, you have two possibilities:
1853 1. Zero event counters before entering the program part you want
1855 to profile, and dump the event counters to a file after leaving
1856 that program part.
1857 2. Switch on/off collection state as needed to only see event
1859 counters happening while inside of the program part you want to
1860 profile.
1861 The second option can be used if the program part you want to
1863 profile is called many times. Option 1, i.e. creating a lot of
1864 dumps is not practical here.
1865 Collection state can be toggled at entry and exit of a given
1867 function with the option --toggle-collect. If you use this option,
1868 collection state should be disabled at the beginning. Note that the
1869 specification of --toggle-collect implicitly sets
1870 --collect-state=no.
1871 Collection state can be toggled also by inserting the client
1873 request CALLGRIND_TOGGLE_COLLECT ; at the needed code positions.
1874 --toggle-collect=<function>
1876 Toggle collection on entry/exit of function.
1877 --collect-jumps=<no|yes> [default: no]
1879 This specifies whether information for (conditional) jumps should
1880 be collected. As above, callgrind_annotate currently is not able to
1881 show you the data. You have to use KCachegrind to get jump arrows
1882 in the annotated code.
1883 --collect-systime=<no|yes> [default: no]
1885 This specifies whether information for system call times should be
1886 collected.
1887 --collect-bus=<no|yes> [default: no]
1889 This specifies whether the number of global bus events executed
1890 should be collected. The event type "Ge" is used for these events.
1891 --cache-sim=<yes|no> [default: no]
1893 Specify if you want to do full cache simulation. By default, only
1894 instruction read accesses will be counted ("Ir"). With cache
1895 simulation, further event counters are enabled: Cache misses on
1896 instruction reads ("I1mr"/"ILmr"), data read accesses ("Dr") and
1897 related cache misses ("D1mr"/"DLmr"), data write accesses ("Dw")
1898 and related cache misses ("D1mw"/"DLmw"). For more information, see
1899 Cachegrind: a cache and branch-prediction profiler.
1900 --branch-sim=<yes|no> [default: no]
1902 Specify if you want to do branch prediction simulation. Further
1903 event counters are enabled: Number of executed conditional branches
1904 and related predictor misses ("Bc"/"Bcm"), executed indirect jumps
1905 and related misses of the jump address predictor ("Bi"/"Bim").
1906
1908 --free-is-write=no|yes [default: no]
1909 When enabled (not the default), Helgrind treats freeing of heap
1910 memory as if the memory was written immediately before the free.
1911 This exposes races where memory is referenced by one thread, and
1912 freed by another, but there is no observable synchronisation event
1913 to ensure that the reference happens before the free.
1914 This functionality is new in Valgrind 3.7.0, and is regarded as
1916 experimental. It is not enabled by default because its interaction
1917 with custom memory allocators is not well understood at present.
1918 User feedback is welcomed.
1919 --track-lockorders=no|yes [default: yes]
1921 When enabled (the default), Helgrind performs lock order
1922 consistency checking. For some buggy programs, the large number of
1923 lock order errors reported can become annoying, particularly if
1924 you're only interested in race errors. You may therefore find it
1925 helpful to disable lock order checking.
1926 --history-level=none|approx|full [default: full]
1928 --history-level=full (the default) causes Helgrind collects enough
1929 information about "old" accesses that it can produce two stack
1930 traces in a race report -- both the stack trace for the current
1931 access, and the trace for the older, conflicting access. To limit
1932 memory usage, "old" accesses stack traces are limited to a maximum
1933 of 8 entries, even if --num-callers value is bigger.
1934 Collecting such information is expensive in both speed and memory,
1936 particularly for programs that do many inter-thread synchronisation
1937 events (locks, unlocks, etc). Without such information, it is more
1938 difficult to track down the root causes of races. Nonetheless, you
1939 may not need it in situations where you just want to check for the
1940 presence or absence of races, for example, when doing regression
1941 testing of a previously race-free program.
1942 --history-level=none is the opposite extreme. It causes Helgrind
1944 not to collect any information about previous accesses. This can be
1945 dramatically faster than --history-level=full.
1946 --history-level=approx provides a compromise between these two
1948 extremes. It causes Helgrind to show a full trace for the later
1949 access, and approximate information regarding the earlier access.
1950 This approximate information consists of two stacks, and the
1951 earlier access is guaranteed to have occurred somewhere between
1952 program points denoted by the two stacks. This is not as useful as
1953 showing the exact stack for the previous access (as
1954 --history-level=full does), but it is better than nothing, and it
1955 is almost as fast as --history-level=none.
1956 --delta-stacktrace=no|yes [default: yes on linux amd64/x86]
1958 This flag only has any effect at --history-level=full.
1959 --delta-stacktrace configures the way Helgrind captures the
1961 stacktraces for the option --history-level=full. Such a stacktrace
1962 is typically needed each time a new piece of memory is read or
1963 written in a basic block of instructions.
1964 --delta-stacktrace=no causes Helgrind to compute a full history
1966 stacktrace from the unwind info each time a stacktrace is needed.
1967 --delta-stacktrace=yes indicates to Helgrind to derive a new
1969 stacktrace from the previous stacktrace, as long as there was no
1970 call instruction, no return instruction, or any other instruction
1971 changing the call stack since the previous stacktrace was captured.
1972 If no such instruction was executed, the new stacktrace can be
1973 derived from the previous stacktrace by just changing the top frame
1974 to the current program counter. This option can speed up Helgrind
1975 by 25% when using --history-level=full.
1976 The following aspects have to be considered when using
1978 --delta-stacktrace=yes :
1979 · In some cases (for example in a function prologue), the
1981 valgrind unwinder might not properly unwind the
1982 stack, due to some
1983 limitations and/or due to wrong unwind info. When
1984 using
1985 --delta-stacktrace=yes, the wrong stack trace
1986 captured in the
1987 function prologue will be kept till the next call
1988 or return.
1989 .RE
1990 · On the other hand, --delta-stacktrace=yes sometimes helps
1992 to
1993 obtain a correct stacktrace, for example when
1994 the unwind info allows
1995 a correct stacktrace to be done in the
1996 beginning of the sequence,
1997 but not later on in the instruction
1998 sequence..RE
1999 · Determining which instructions are changing the
2001 callstack is
2002 partially based on platform dependent
2003 heuristics, which have to be
2004 tuned/validated specifically for the
2005 platform. Also, unwinding in a
2006 function prologue must be good enough to
2007 allow using
2008 --delta-stacktrace=yes. Currently, the
2009 option --delta-stacktrace=yes
2010 has been reasonably validated only on linux
2011 x86 32 bits and linux
2012 amd64 64 bits. For more details about how
2013 to validate
2014 --delta-stacktrace=yes, see debug option
2015 --hg-sanity-flags and the
2016 function check_cached_rcec_ok in
2017 libhb_core.c..RE
2018 --conflict-cache-size=N [default: 1000000]
2021 This flag only has any effect at --history-level=full.
2022 Information about "old" conflicting accesses is stored
2024 in a cache of limited size, with LRU-style management.
2025 This is necessary because it isn't practical to store a
2026 stack trace for every single memory access made by the
2027 program. Historical information on not recently
2028 accessed locations is periodically discarded, to free
2029 up space in the cache.
2030 This option controls the size of the cache, in terms of
2032 the number of different memory addresses for which
2033 conflicting access information is stored. If you find
2034 that Helgrind is showing race errors with only one
2035 stack instead of the expected two stacks, try
2036 increasing this value.
2037 The minimum value is 10,000 and the maximum is
2039 30,000,000 (thirty times the default value). Increasing
2040 the value by 1 increases Helgrind's memory requirement
2041 by very roughly 100 bytes, so the maximum value will
2042 easily eat up three extra gigabytes or so of memory.
2043 --check-stack-refs=no|yes [default: yes]
2045 By default Helgrind checks all data memory accesses
2046 made by your program. This flag enables you to skip
2047 checking for accesses to thread stacks (local
2048 variables). This can improve performance, but comes at
2049 the cost of missing races on stack-allocated data.
2050 --ignore-thread-creation=<yes|no> [default: no]
2052 Controls whether all activities during thread creation
2053 should be ignored. By default enabled only on Solaris.
2054 Solaris provides higher throughput, parallelism and
2055 scalability than other operating systems, at the cost
2056 of more fine-grained locking activity. This means for
2057 example that when a thread is created under glibc, just
2058 one big lock is used for all thread setup. Solaris libc
2059 uses several fine-grained locks and the creator thread
2060 resumes its activities as soon as possible, leaving for
2061 example stack and TLS setup sequence to the created
2062 thread. This situation confuses Helgrind as it assumes
2063 there is some false ordering in place between creator
2064 and created thread; and therefore many types of race
2065 conditions in the application would not be reported. To
2066 prevent such false ordering, this command line option
2067 is set to yes by default on Solaris. All activity
2068 (loads, stores, client requests) is therefore ignored
2069 during:
2070 · pthread_create() call in the creator thread
2072 · thread creation phase (stack and TLS setup) in the
2074 created thread
2075 Also new memory allocated during thread creation is
2077 untracked, that is race reporting is suppressed there.
2078 DRD does the same thing implicitly. This is necessary
2079 because Solaris libc caches many objects and reuses
2080 them for different threads and that confuses Helgrind.
2081
2083 --check-stack-var=<yes|no> [default: no]
2084 Controls whether DRD detects data races on stack variables.
2085 Verifying stack variables is disabled by default because most
2086 programs do not share stack variables over threads.
2087 --exclusive-threshold=<n> [default: off]
2089 Print an error message if any mutex or writer lock has been held
2090 longer than the time specified in milliseconds. This option enables
2091 the detection of lock contention.
2092 --join-list-vol=<n> [default: 10]
2094 Data races that occur between a statement at the end of one thread
2095 and another thread can be missed if memory access information is
2096 discarded immediately after a thread has been joined. This option
2097 allows to specify for how many joined threads memory access
2098 information should be retained.
2099 --first-race-only=<yes|no> [default: no]
2101 Whether to report only the first data race that has been detected
2102 on a memory location or all data races that have been detected on a
2103 memory location.
2104 --free-is-write=<yes|no> [default: no]
2106 Whether to report races between accessing memory and freeing
2107 memory. Enabling this option may cause DRD to run slightly slower.
2108 Notes:
2109 · Don't enable this option when using custom memory allocators
2111 that use the VG_USERREQ__MALLOCLIKE_BLOCK and
2112 VG_USERREQ__FREELIKE_BLOCK because that would result in false
2113 positives.
2114 · Don't enable this option when using reference-counted objects
2116 because that will result in false positives, even when that
2117 code has been annotated properly with ANNOTATE_HAPPENS_BEFORE
2118 and ANNOTATE_HAPPENS_AFTER. See e.g. the output of the
2119 following command for an example: valgrind --tool=drd
2120 --free-is-write=yes drd/tests/annotate_smart_pointer.
2121 --report-signal-unlocked=<yes|no> [default: yes]
2123 Whether to report calls to pthread_cond_signal and
2124 pthread_cond_broadcast where the mutex associated with the signal
2125 through pthread_cond_wait or pthread_cond_timed_waitis not locked
2126 at the time the signal is sent. Sending a signal without holding a
2127 lock on the associated mutex is a common programming error which
2128 can cause subtle race conditions and unpredictable behavior. There
2129 exist some uncommon synchronization patterns however where it is
2130 safe to send a signal without holding a lock on the associated
2131 mutex.
2132 --segment-merging=<yes|no> [default: yes]
2134 Controls segment merging. Segment merging is an algorithm to limit
2135 memory usage of the data race detection algorithm. Disabling
2136 segment merging may improve the accuracy of the so-called 'other
2137 segments' displayed in race reports but can also trigger an out of
2138 memory error.
2139 --segment-merging-interval=<n> [default: 10]
2141 Perform segment merging only after the specified number of new
2142 segments have been created. This is an advanced configuration
2143 option that allows to choose whether to minimize DRD's memory usage
2144 by choosing a low value or to let DRD run faster by choosing a
2145 slightly higher value. The optimal value for this parameter depends
2146 on the program being analyzed. The default value works well for
2147 most programs.
2148 --shared-threshold=<n> [default: off]
2150 Print an error message if a reader lock has been held longer than
2151 the specified time (in milliseconds). This option enables the
2152 detection of lock contention.
2153 --show-confl-seg=<yes|no> [default: yes]
2155 Show conflicting segments in race reports. Since this information
2156 can help to find the cause of a data race, this option is enabled
2157 by default. Disabling this option makes the output of DRD more
2158 compact.
2159 --show-stack-usage=<yes|no> [default: no]
2161 Print stack usage at thread exit time. When a program creates a
2162 large number of threads it becomes important to limit the amount of
2163 virtual memory allocated for thread stacks. This option makes it
2164 possible to observe how much stack memory has been used by each
2165 thread of the client program. Note: the DRD tool itself allocates
2166 some temporary data on the client thread stack. The space necessary
2167 for this temporary data must be allocated by the client program
2168 when it allocates stack memory, but is not included in stack usage
2169 reported by DRD.
2170 --ignore-thread-creation=<yes|no> [default: no]
2172 Controls whether all activities during thread creation should be
2173 ignored. By default enabled only on Solaris. Solaris provides
2174 higher throughput, parallelism and scalability than other operating
2175 systems, at the cost of more fine-grained locking activity. This
2176 means for example that when a thread is created under glibc, just
2177 one big lock is used for all thread setup. Solaris libc uses
2178 several fine-grained locks and the creator thread resumes its
2179 activities as soon as possible, leaving for example stack and TLS
2180 setup sequence to the created thread. This situation confuses DRD
2181 as it assumes there is some false ordering in place between creator
2182 and created thread; and therefore many types of race conditions in
2183 the application would not be reported. To prevent such false
2184 ordering, this command line option is set to yes by default on
2185 Solaris. All activity (loads, stores, client requests) is therefore
2186 ignored during:
2187 · pthread_create() call in the creator thread
2189 · thread creation phase (stack and TLS setup) in the created
2191 thread
2192 --trace-addr=<address> [default: none]
2194 Trace all load and store activity for the specified address. This
2195 option may be specified more than once.
2196 --ptrace-addr=<address> [default: none]
2198 Trace all load and store activity for the specified address and
2199 keep doing that even after the memory at that address has been
2200 freed and reallocated.
2201 --trace-alloc=<yes|no> [default: no]
2203 Trace all memory allocations and deallocations. May produce a huge
2204 amount of output.
2205 --trace-barrier=<yes|no> [default: no]
2207 Trace all barrier activity.
2208 --trace-cond=<yes|no> [default: no]
2210 Trace all condition variable activity.
2211 --trace-fork-join=<yes|no> [default: no]
2213 Trace all thread creation and all thread termination events.
2214 --trace-hb=<yes|no> [default: no]
2216 Trace execution of the ANNOTATE_HAPPENS_BEFORE(),
2217 ANNOTATE_HAPPENS_AFTER() and ANNOTATE_HAPPENS_DONE() client
2218 requests.
2219 --trace-mutex=<yes|no> [default: no]
2221 Trace all mutex activity.
2222 --trace-rwlock=<yes|no> [default: no]
2224 Trace all reader-writer lock activity.
2225 --trace-semaphore=<yes|no> [default: no]
2227 Trace all semaphore activity.
2228
2230 --heap=<yes|no> [default: yes]
2231 Specifies whether heap profiling should be done.
2232 --heap-admin=<size> [default: 8]
2234 If heap profiling is enabled, gives the number of administrative
2235 bytes per block to use. This should be an estimate of the average,
2236 since it may vary. For example, the allocator used by glibc on
2237 Linux requires somewhere between 4 to 15 bytes per block, depending
2238 on various factors. That allocator also requires admin space for
2239 freed blocks, but Massif cannot account for this.
2240 --stacks=<yes|no> [default: no]
2242 Specifies whether stack profiling should be done. This option slows
2243 Massif down greatly, and so is off by default. Note that Massif
2244 assumes that the main stack has size zero at start-up. This is not
2245 true, but doing otherwise accurately is difficult. Furthermore,
2246 starting at zero better indicates the size of the part of the main
2247 stack that a user program actually has control over.
2248 --pages-as-heap=<yes|no> [default: no]
2250 Tells Massif to profile memory at the page level rather than at the
2251 malloc'd block level. See above for details.
2252 --depth=<number> [default: 30]
2254 Maximum depth of the allocation trees recorded for detailed
2255 snapshots. Increasing it will make Massif run somewhat more slowly,
2256 use more memory, and produce bigger output files.
2257 --alloc-fn=<name>
2259 Functions specified with this option will be treated as though they
2260 were a heap allocation function such as malloc. This is useful for
2261 functions that are wrappers to malloc or new, which can fill up the
2262 allocation trees with uninteresting information. This option can be
2263 specified multiple times on the command line, to name multiple
2264 functions.
2265 Note that the named function will only be treated this way if it is
2267 the top entry in a stack trace, or just below another function
2268 treated this way. For example, if you have a function malloc1 that
2269 wraps malloc, and malloc2 that wraps malloc1, just specifying
2270 --alloc-fn=malloc2 will have no effect. You need to specify
2271 --alloc-fn=malloc1 as well. This is a little inconvenient, but the
2272 reason is that checking for allocation functions is slow, and it
2273 saves a lot of time if Massif can stop looking through the stack
2274 trace entries as soon as it finds one that doesn't match rather
2275 than having to continue through all the entries.
2276 Note that C++ names are demangled. Note also that overloaded C++
2278 names must be written in full. Single quotes may be necessary to
2279 prevent the shell from breaking them up. For example:
2280 --alloc-fn='operator new(unsigned, std::nothrow_t const&)'
2282 --ignore-fn=<name>
2285 Any direct heap allocation (i.e. a call to malloc, new, etc, or a
2286 call to a function named by an --alloc-fn option) that occurs in a
2287 function specified by this option will be ignored. This is mostly
2288 useful for testing purposes. This option can be specified multiple
2289 times on the command line, to name multiple functions.
2290 Any realloc of an ignored block will also be ignored, even if the
2292 realloc call does not occur in an ignored function. This avoids the
2293 possibility of negative heap sizes if ignored blocks are shrunk
2294 with realloc.
2295 The rules for writing C++ function names are the same as for
2297 --alloc-fn above.
2298 --threshold=<m.n> [default: 1.0]
2300 The significance threshold for heap allocations, as a percentage of
2301 total memory size. Allocation tree entries that account for less
2302 than this will be aggregated. Note that this should be specified in
2303 tandem with ms_print's option of the same name.
2304 --peak-inaccuracy=<m.n> [default: 1.0]
2306 Massif does not necessarily record the actual global memory
2307 allocation peak; by default it records a peak only when the global
2308 memory allocation size exceeds the previous peak by at least 1.0%.
2309 This is because there can be many local allocation peaks along the
2310 way, and doing a detailed snapshot for every one would be expensive
2311 and wasteful, as all but one of them will be later discarded. This
2312 inaccuracy can be changed (even to 0.0%) via this option, but
2313 Massif will run drastically slower as the number approaches zero.
2314 --time-unit=<i|ms|B> [default: i]
2316 The time unit used for the profiling. There are three
2317 possibilities: instructions executed (i), which is good for most
2318 cases; real (wallclock) time (ms, i.e. milliseconds), which is
2319 sometimes useful; and bytes allocated/deallocated on the heap
2320 and/or stack (B), which is useful for very short-run programs, and
2321 for testing purposes, because it is the most reproducible across
2322 different machines.
2323 --detailed-freq=<n> [default: 10]
2325 Frequency of detailed snapshots. With --detailed-freq=1, every
2326 snapshot is detailed.
2327 --max-snapshots=<n> [default: 100]
2329 The maximum number of snapshots recorded. If set to N, for all
2330 programs except very short-running ones, the final number of
2331 snapshots will be between N/2 and N.
2332 --massif-out-file=<file> [default: massif.out.%p]
2334 Write the profile data to file rather than to the default output
2335 file, massif.out.<pid>. The %p and %q format specifiers can be used
2336 to embed the process ID and/or the contents of an environment
2337 variable in the name, as is the case for the core option --log-
2338 file.
2339
2341 There are no SGCheck-specific command-line options at present.
2342
2344 --bb-out-file=<name> [default: bb.out.%p]
2345 This option selects the name of the basic block vector file. The %p
2346 and %q format specifiers can be used to embed the process ID and/or
2347 the contents of an environment variable in the name, as is the case
2348 for the core option --log-file.
2349 --pc-out-file=<name> [default: pc.out.%p]
2351 This option selects the name of the PC file. This file holds
2352 program counter addresses and function name info for the various
2353 basic blocks. This can be used in conjunction with the basic block
2354 vector file to fast-forward via function names instead of just
2355 instruction counts. The %p and %q format specifiers can be used to
2356 embed the process ID and/or the contents of an environment variable
2357 in the name, as is the case for the core option --log-file.
2358 --interval-size=<number> [default: 100000000]
2360 This option selects the size of the interval to use. The default is
2361 100 million instructions, which is a commonly used value. Other
2362 sizes can be used; smaller intervals can help programs with
2363 finer-grained phases. However smaller interval size can lead to
2364 accuracy issues due to warm-up effects (When fast-forwarding the
2365 various architectural features will be un-initialized, and it will
2366 take some number of instructions before they "warm up" to the state
2367 a full simulation would be at without the fast-forwarding. Large
2368 interval sizes tend to mitigate this.)
2369 --instr-count-only [default: no]
2371 This option tells the tool to only display instruction count
2372 totals, and to not generate the actual basic block vector file.
2373 This is useful for debugging, and for gathering instruction count
2374 info without generating the large basic block vector files.
2375
2377 --basic-counts=<no|yes> [default: yes]
2378 When enabled, Lackey prints the following statistics and
2379 information about the execution of the client program:
2380 1. The number of calls to the function specified by the --fnname
2382 option (the default is main). If the program has had its
2383 symbols stripped, the count will always be zero.
2384 2. The number of conditional branches encountered and the number
2386 and proportion of those taken.
2387 3. The number of superblocks entered and completed by the program.
2389 Note that due to optimisations done by the JIT, this is not at
2390 all an accurate value.
2391 4. The number of guest (x86, amd64, ppc, etc.) instructions and IR
2393 statements executed. IR is Valgrind's RISC-like intermediate
2394 representation via which all instrumentation is done.
2395 5. Ratios between some of these counts.
2397 6. The exit code of the client program.
2399 --detailed-counts=<no|yes> [default: no]
2401 When enabled, Lackey prints a table containing counts of loads,
2402 stores and ALU operations, differentiated by their IR types. The IR
2403 types are identified by their IR name ("I1", "I8", ... "I128",
2404 "F32", "F64", and "V128").
2405 --trace-mem=<no|yes> [default: no]
2407 When enabled, Lackey prints the size and address of almost every
2408 memory access made by the program. See the comments at the top of
2409 the file lackey/lk_main.c for details about the output format, how
2410 it works, and inaccuracies in the address trace. Note that this
2411 option produces immense amounts of output.
2412 --trace-superblocks=<no|yes> [default: no]
2414 When enabled, Lackey prints out the address of every superblock (a
2415 single entry, multiple exit, linear chunk of code) executed by the
2416 program. This is primarily of interest to Valgrind developers. See
2417 the comments at the top of the file lackey/lk_main.c for details
2418 about the output format. Note that this option produces large
2419 amounts of output.
2420 --fnname=<name> [default: main]
2422 Changes the function for which calls are counted when
2423 --basic-counts=yes is specified.
2424
2426 cg_annotate(1), callgrind_annotate(1), callgrind_control(1),
2427 ms_print(1), $INSTALL/share/doc/valgrind/html/index.html or
2428 http://www.valgrind.org/docs/manual/index.html, Debugging your program
2429 using Valgrind's gdbserver and GDB[1] vgdb[2], Valgrind monitor
2430 commands[3], The Commentary[4], Scheduling and Multi-Thread
2431 Performance[5], Cachegrind: a cache and branch-prediction profiler[6].
2432
2434 See the AUTHORS file in the valgrind distribution for a comprehensive
2435 list of authors.
2436 This manpage was written by Andres Roldan <aroldan@debian.org> and the
2438 Valgrind developers.
2439
2441 1. Debugging your program using Valgrind's gdbserver and GDB
2442 http://www.valgrind.org/docs/manual/manual-core-adv.html#manual-core-adv.gdbserver
2443 2. vgdb
2445 http://www.valgrind.org/docs/manual/manual-core-adv.html#manual-core-adv.vgdb
2446 3. Valgrind monitor commands
2448 http://www.valgrind.org/docs/manual/manual-core-adv.html#manual-core-adv.valgrind-monitor-commands
2449 4. The Commentary
2451 http://www.valgrind.org/docs/manual/manual-core.html#manual-core.comment
2452 5. Scheduling and Multi-Thread Performance
2454 http://www.valgrind.org/docs/manual/manual-core.html#manual-core.pthreads_perf_sched
2455 6. Cachegrind: a cache and branch-prediction profiler
2457 http://www.valgrind.org/docs/manual/cg-manual.html
2458Release 3.15.0 04/13/2019 VALGRIND(1)