1VALGRIND(1) Release 3.14.0 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, lackey, none, exp-sgcheck,
33 exp-bbv, exp-dhat, 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 --sigill-diagnostics=<yes|no> [default: yes]
393 Enable/disable printing of illegal instruction diagnostics. Enabled
394 by default, but defaults to disabled when --quiet is given. The
395 default can always be explicitly overridden by giving this option.
396 When enabled, a warning message will be printed, along with some
398 diagnostics, whenever an instruction is encountered that Valgrind
399 cannot decode or translate, before the program is given a SIGILL
400 signal. Often an illegal instruction indicates a bug in the program
401 or missing support for the particular instruction in Valgrind. But
402 some programs do deliberately try to execute an instruction that
403 might be missing and trap the SIGILL signal to detect processor
404 features. Using this flag makes it possible to avoid the diagnostic
405 output that you would otherwise get in such cases.
406 --keep-debuginfo=<yes|no> [default: no]
408 When enabled, keep ("archive") symbols and all other debuginfo for
409 unloaded code. This allows saved stack traces to include file/line
410 info for code that has been dlclose'd (or similar). Be careful with
411 this, since it can lead to unbounded memory use for programs which
412 repeatedly load and unload shared objects.
413 Some tools and some functionalities have only limited support for
415 archived debug info. Memcheck fully supports it. Generally, tools
416 that report errors can use archived debug info to show the error
417 stack traces. The known limitations are: Helgrind's past access
418 stack trace of a race condition is does not use archived debug
419 info. Massif (and more generally the xtree Massif output format)
420 does not make use of archived debug info. Only Memcheck has been
421 (somewhat) tested with --keep-debuginfo=yes, so other tools may
422 have unknown limitations.
423 --show-below-main=<yes|no> [default: no]
425 By default, stack traces for errors do not show any functions that
426 appear beneath main because most of the time it's uninteresting C
427 library stuff and/or gobbledygook. Alternatively, if main is not
428 present in the stack trace, stack traces will not show any
429 functions below main-like functions such as glibc's
430 __libc_start_main. Furthermore, if main-like functions are present
431 in the trace, they are normalised as (below main), in order to make
432 the output more deterministic.
433 If this option is enabled, all stack trace entries will be shown
435 and main-like functions will not be normalised.
436 --fullpath-after=<string> [default: don't show source paths]
438 By default Valgrind only shows the filenames in stack traces, but
439 not full paths to source files. When using Valgrind in large
440 projects where the sources reside in multiple different
441 directories, this can be inconvenient. --fullpath-after provides a
442 flexible solution to this problem. When this option is present, the
443 path to each source file is shown, with the following all-important
444 caveat: if string is found in the path, then the path up to and
445 including string is omitted, else the path is shown unmodified.
446 Note that string is not required to be a prefix of the path.
447 For example, consider a file named
449 /home/janedoe/blah/src/foo/bar/xyzzy.c. Specifying
450 --fullpath-after=/home/janedoe/blah/src/ will cause Valgrind to
451 show the name as foo/bar/xyzzy.c.
452 Because the string is not required to be a prefix,
454 --fullpath-after=src/ will produce the same output. This is useful
455 when the path contains arbitrary machine-generated characters. For
456 example, the path /my/build/dir/C32A1B47/blah/src/foo/xyzzy can be
457 pruned to foo/xyzzy using --fullpath-after=/blah/src/.
458 If you simply want to see the full path, just specify an empty
460 string: --fullpath-after=. This isn't a special case, merely a
461 logical consequence of the above rules.
462 Finally, you can use --fullpath-after multiple times. Any
464 appearance of it causes Valgrind to switch to producing full paths
465 and applying the above filtering rule. Each produced path is
466 compared against all the --fullpath-after-specified strings, in the
467 order specified. The first string to match causes the path to be
468 truncated as described above. If none match, the full path is
469 shown. This facilitates chopping off prefixes when the sources are
470 drawn from a number of unrelated directories.
471 --extra-debuginfo-path=<path> [default: undefined and unused]
473 By default Valgrind searches in several well-known paths for debug
474 objects, such as /usr/lib/debug/.
475 However, there may be scenarios where you may wish to put debug
477 objects at an arbitrary location, such as external storage when
478 running Valgrind on a mobile device with limited local storage.
479 Another example might be a situation where you do not have
480 permission to install debug object packages on the system where you
481 are running Valgrind.
482 In these scenarios, you may provide an absolute path as an extra,
484 final place for Valgrind to search for debug objects by specifying
485 --extra-debuginfo-path=/path/to/debug/objects. The given path will
486 be prepended to the absolute path name of the searched-for object.
487 For example, if Valgrind is looking for the debuginfo for
488 /w/x/y/zz.so and --extra-debuginfo-path=/a/b/c is specified, it
489 will look for a debug object at /a/b/c/w/x/y/zz.so.
490 This flag should only be specified once. If it is specified
492 multiple times, only the last instance is honoured.
493 --debuginfo-server=ipaddr:port [default: undefined and unused]
495 This is a new, experimental, feature introduced in version 3.9.0.
496 In some scenarios it may be convenient to read debuginfo from
498 objects stored on a different machine. With this flag, Valgrind
499 will query a debuginfo server running on ipaddr and listening on
500 port port, if it cannot find the debuginfo object in the local
501 filesystem.
502 The debuginfo server must accept TCP connections on port port. The
504 debuginfo server is contained in the source file
505 auxprogs/valgrind-di-server.c. It will only serve from the
506 directory it is started in. port defaults to 1500 in both client
507 and server if not specified.
508 If Valgrind looks for the debuginfo for /w/x/y/zz.so by using the
510 debuginfo server, it will strip the pathname components and merely
511 request zz.so on the server. That in turn will look only in its
512 current working directory for a matching debuginfo object.
513 The debuginfo data is transmitted in small fragments (8 KB) as
515 requested by Valgrind. Each block is compressed using LZO to reduce
516 transmission time. The implementation has been tuned for best
517 performance over a single-stage 802.11g (WiFi) network link.
518 Note that checks for matching primary vs debug objects, using GNU
520 debuglink CRC scheme, are performed even when using the debuginfo
521 server. To disable such checking, you need to also specify
522 --allow-mismatched-debuginfo=yes.
523 By default the Valgrind build system will build valgrind-di-server
525 for the target platform, which is almost certainly not what you
526 want. So far we have been unable to find out how to get
527 automake/autoconf to build it for the build platform. If you want
528 to use it, you will have to recompile it by hand using the command
529 shown at the top of auxprogs/valgrind-di-server.c.
530 --allow-mismatched-debuginfo=no|yes [no]
532 When reading debuginfo from separate debuginfo objects, Valgrind
533 will by default check that the main and debuginfo objects match,
534 using the GNU debuglink mechanism. This guarantees that it does not
535 read debuginfo from out of date debuginfo objects, and also ensures
536 that Valgrind can't crash as a result of mismatches.
537 This check can be overridden using
539 --allow-mismatched-debuginfo=yes. This may be useful when the
540 debuginfo and main objects have not been split in the proper way.
541 Be careful when using this, though: it disables all consistency
542 checking, and Valgrind has been observed to crash when the main and
543 debuginfo objects don't match.
544 --suppressions=<filename> [default: $PREFIX/lib/valgrind/default.supp]
546 Specifies an extra file from which to read descriptions of errors
547 to suppress. You may use up to 100 extra suppression files.
548 --gen-suppressions=<yes|no|all> [default: no]
550 When set to yes, Valgrind will pause after every error shown and
551 print the line:
552 ---- Print suppression ? --- [Return/N/n/Y/y/C/c] ----
554 Pressing Ret, or N Ret or n Ret, causes Valgrind continue execution
556 without printing a suppression for this error.
557 Pressing Y Ret or y Ret causes Valgrind to write a suppression for
559 this error. You can then cut and paste it into a suppression file
560 if you don't want to hear about the error in the future.
561 When set to all, Valgrind will print a suppression for every
563 reported error, without querying the user.
564 This option is particularly useful with C++ programs, as it prints
566 out the suppressions with mangled names, as required.
567 Note that the suppressions printed are as specific as possible. You
569 may want to common up similar ones, by adding wildcards to function
570 names, and by using frame-level wildcards. The wildcarding
571 facilities are powerful yet flexible, and with a bit of careful
572 editing, you may be able to suppress a whole family of related
573 errors with only a few suppressions.
574 Sometimes two different errors are suppressed by the same
576 suppression, in which case Valgrind will output the suppression
577 more than once, but you only need to have one copy in your
578 suppression file (but having more than one won't cause problems).
579 Also, the suppression name is given as <insert a suppression name
580 here>; the name doesn't really matter, it's only used with the -v
581 option which prints out all used suppression records.
582 --input-fd=<number> [default: 0, stdin]
584 When using --gen-suppressions=yes, Valgrind will stop so as to read
585 keyboard input from you when each error occurs. By default it reads
586 from the standard input (stdin), which is problematic for programs
587 which close stdin. This option allows you to specify an alternative
588 file descriptor from which to read input.
589 --dsymutil=no|yes [yes]
591 This option is only relevant when running Valgrind on Mac OS X.
592 Mac OS X uses a deferred debug information (debuginfo) linking
594 scheme. When object files containing debuginfo are linked into a
595 .dylib or an executable, the debuginfo is not copied into the final
596 file. Instead, the debuginfo must be linked manually by running
597 dsymutil, a system-provided utility, on the executable or .dylib.
598 The resulting combined debuginfo is placed in a directory alongside
599 the executable or .dylib, but with the extension .dSYM.
600 With --dsymutil=no, Valgrind will detect cases where the .dSYM
602 directory is either missing, or is present but does not appear to
603 match the associated executable or .dylib, most likely because it
604 is out of date. In these cases, Valgrind will print a warning
605 message but take no further action.
606 With --dsymutil=yes, Valgrind will, in such cases, automatically
608 run dsymutil as necessary to bring the debuginfo up to date. For
609 all practical purposes, if you always use --dsymutil=yes, then
610 there is never any need to run dsymutil manually or as part of your
611 applications's build system, since Valgrind will run it as
612 necessary.
613 Valgrind will not attempt to run dsymutil on any executable or
615 library in /usr/, /bin/, /sbin/, /opt/, /sw/, /System/, /Library/
616 or /Applications/ since dsymutil will always fail in such
617 situations. It fails both because the debuginfo for such
618 pre-installed system components is not available anywhere, and also
619 because it would require write privileges in those directories.
620 Be careful when using --dsymutil=yes, since it will cause
622 pre-existing .dSYM directories to be silently deleted and
623 re-created. Also note that dsymutil is quite slow, sometimes
624 excessively so.
625 --max-stackframe=<number> [default: 2000000]
627 The maximum size of a stack frame. If the stack pointer moves by
628 more than this amount then Valgrind will assume that the program is
629 switching to a different stack.
630 You may need to use this option if your program has large
632 stack-allocated arrays. Valgrind keeps track of your program's
633 stack pointer. If it changes by more than the threshold amount,
634 Valgrind assumes your program is switching to a different stack,
635 and Memcheck behaves differently than it would for a stack pointer
636 change smaller than the threshold. Usually this heuristic works
637 well. However, if your program allocates large structures on the
638 stack, this heuristic will be fooled, and Memcheck will
639 subsequently report large numbers of invalid stack accesses. This
640 option allows you to change the threshold to a different value.
641 You should only consider use of this option if Valgrind's debug
643 output directs you to do so. In that case it will tell you the new
644 threshold you should specify.
645 In general, allocating large structures on the stack is a bad idea,
647 because you can easily run out of stack space, especially on
648 systems with limited memory or which expect to support large
649 numbers of threads each with a small stack, and also because the
650 error checking performed by Memcheck is more effective for
651 heap-allocated data than for stack-allocated data. If you have to
652 use this option, you may wish to consider rewriting your code to
653 allocate on the heap rather than on the stack.
654 --main-stacksize=<number> [default: use current 'ulimit' value]
656 Specifies the size of the main thread's stack.
657 To simplify its memory management, Valgrind reserves all required
659 space for the main thread's stack at startup. That means it needs
660 to know the required stack size at startup.
661 By default, Valgrind uses the current "ulimit" value for the stack
663 size, or 16 MB, whichever is lower. In many cases this gives a
664 stack size in the range 8 to 16 MB, which almost never overflows
665 for most applications.
666 If you need a larger total stack size, use --main-stacksize to
668 specify it. Only set it as high as you need, since reserving far
669 more space than you need (that is, hundreds of megabytes more than
670 you need) constrains Valgrind's memory allocators and may reduce
671 the total amount of memory that Valgrind can use. This is only
672 really of significance on 32-bit machines.
673 On Linux, you may request a stack of size up to 2GB. Valgrind will
675 stop with a diagnostic message if the stack cannot be allocated.
676 --main-stacksize only affects the stack size for the program's
678 initial thread. It has no bearing on the size of thread stacks, as
679 Valgrind does not allocate those.
680 You may need to use both --main-stacksize and --max-stackframe
682 together. It is important to understand that --main-stacksize sets
683 the maximum total stack size, whilst --max-stackframe specifies the
684 largest size of any one stack frame. You will have to work out the
685 --main-stacksize value for yourself (usually, if your applications
686 segfaults). But Valgrind will tell you the needed --max-stackframe
687 size, if necessary.
688 As discussed further in the description of --max-stackframe, a
690 requirement for a large stack is a sign of potential portability
691 problems. You are best advised to place all large data in
692 heap-allocated memory.
693 --max-threads=<number> [default: 500]
695 By default, Valgrind can handle to up to 500 threads. Occasionally,
696 that number is too small. Use this option to provide a different
697 limit. E.g. --max-threads=3000.
698
700 For tools that use their own version of malloc (e.g. Memcheck, Massif,
701 Helgrind, DRD), the following options apply.
702 --alignment=<number> [default: 8 or 16, depending on the platform]
704 By default Valgrind's malloc, realloc, etc, return a block whose
705 starting address is 8-byte aligned or 16-byte aligned (the value
706 depends on the platform and matches the platform default). This
707 option allows you to specify a different alignment. The supplied
708 value must be greater than or equal to the default, less than or
709 equal to 4096, and must be a power of two.
710 --redzone-size=<number> [default: depends on the tool]
712 Valgrind's malloc, realloc, etc, add padding blocks before and
713 after each heap block allocated by the program being run. Such
714 padding blocks are called redzones. The default value for the
715 redzone size depends on the tool. For example, Memcheck adds and
716 protects a minimum of 16 bytes before and after each block
717 allocated by the client. This allows it to detect block underruns
718 or overruns of up to 16 bytes.
719 Increasing the redzone size makes it possible to detect overruns of
721 larger distances, but increases the amount of memory used by
722 Valgrind. Decreasing the redzone size will reduce the memory needed
723 by Valgrind but also reduces the chances of detecting
724 over/underruns, so is not recommended.
725 --xtree-memory=none|allocs|full [none]
727 Tools replacing Valgrind's malloc, realloc, etc, can optionally
728 produce an execution tree detailing which piece of code is
729 responsible for heap memory usage. See ??? for a detailed
730 explanation about execution trees.
731 When set to none, no memory execution tree is produced.
733 When set to allocs, the memory execution tree gives the current
735 number of allocated bytes and the current number of allocated
736 blocks.
737 When set to full, the memory execution tree gives 6 different
739 measurements : the current number of allocated bytes and blocks
740 (same values as for allocs), the total number of allocated bytes
741 and blocks, the total number of freed bytes and blocks.
742 Note that the overhead in cpu and memory to produce an xtree
744 depends on the tool. The overhead in cpu is small for the value
745 allocs, as the information needed to produce this report is
746 maintained in any case by the tool. For massif and helgrind,
747 specifying full implies to capture a stack trace for each free
748 operation, while normally these tools only capture an allocation
749 stack trace. For Memcheck, the cpu overhead for the value full is
750 small, as this can only be used in combination with
751 --keep-stacktraces=alloc-and-free or
752 --keep-stacktraces=alloc-then-free, which already records a stack
753 trace for each free operation. The memory overhead varies between 5
754 and 10 words per unique stacktrace in the xtree, plus the memory
755 needed to record the stack trace for the free operations, if needed
756 specifically for the xtree.
757 --xtree-memory-file=<filename> [default: xtmemory.kcg.%p]
759 Specifies that Valgrind should produce the xtree memory report in
760 the specified file. Any %p or %q sequences appearing in the
761 filename are expanded in exactly the same way as they are for
762 --log-file. See the description of --log-file for details.
763 If the filename contains the extension .ms, then the produced file
765 format will be a massif output file format. If the filename
766 contains the extension .kcg or no extension is provided or
767 recognised, then the produced file format will be a callgrind
768 output format.
769 See ??? for a detailed explanation about execution trees formats.
771
773 These options apply to all tools, as they affect certain obscure
774 workings of the Valgrind core. Most people won't need to use them.
775 --smc-check=<none|stack|all|all-non-file> [default: all-non-file for
777 x86/amd64/s390x, stack for other archs]
778 This option controls Valgrind's detection of self-modifying code.
779 If no checking is done, when a program executes some code, then
780 overwrites it with new code, and executes the new code, Valgrind
781 will continue to execute the translations it made for the old code.
782 This will likely lead to incorrect behaviour and/or crashes.
783 For "modern" architectures -- anything that's not x86, amd64 or
785 s390x -- the default is stack. This is because a correct program
786 must take explicit action to reestablish D-I cache coherence
787 following code modification. Valgrind observes and honours such
788 actions, with the result that self-modifying code is transparently
789 handled with zero extra cost.
790 For x86, amd64 and s390x, the program is not required to notify the
792 hardware of required D-I coherence syncing. Hence the default is
793 all-non-file, which covers the normal case of generating code into
794 an anonymous (non-file-backed) mmap'd area.
795 The meanings of the four available settings are as follows. No
797 detection (none), detect self-modifying code on the stack (which is
798 used by GCC to implement nested functions) (stack), detect
799 self-modifying code everywhere (all), and detect self-modifying
800 code everywhere except in file-backed mappings (all-non-file).
801 Running with all will slow Valgrind down noticeably. Running with
803 none will rarely speed things up, since very little code gets
804 dynamically generated in most programs. The
805 VALGRIND_DISCARD_TRANSLATIONS client request is an alternative to
806 --smc-check=all and --smc-check=all-non-file that requires more
807 programmer effort but allows Valgrind to run your program faster,
808 by telling it precisely when translations need to be re-made.
809 --smc-check=all-non-file provides a cheaper but more limited
811 version of --smc-check=all. It adds checks to any translations that
812 do not originate from file-backed memory mappings. Typical
813 applications that generate code, for example JITs in web browsers,
814 generate code into anonymous mmaped areas, whereas the "fixed" code
815 of the browser always lives in file-backed mappings.
816 --smc-check=all-non-file takes advantage of this observation,
817 limiting the overhead of checking to code which is likely to be JIT
818 generated.
819 --read-inline-info=<yes|no> [default: see below]
821 When enabled, Valgrind will read information about inlined function
822 calls from DWARF3 debug info. This slows Valgrind startup and makes
823 it use more memory (typically for each inlined piece of code, 6
824 words and space for the function name), but it results in more
825 descriptive stacktraces. For the 3.10.0 release, this functionality
826 is enabled by default only for Linux, Android and Solaris targets
827 and only for the tools Memcheck, Helgrind and DRD. Here is an
828 example of some stacktraces with --read-inline-info=no:
829 ==15380== Conditional jump or move depends on uninitialised value(s)
831 ==15380== at 0x80484EA: main (inlinfo.c:6)
832 ==15380==
833 ==15380== Conditional jump or move depends on uninitialised value(s)
834 ==15380== at 0x8048550: fun_noninline (inlinfo.c:6)
835 ==15380== by 0x804850E: main (inlinfo.c:34)
836 ==15380==
837 ==15380== Conditional jump or move depends on uninitialised value(s)
838 ==15380== at 0x8048520: main (inlinfo.c:6)
839 And here are the same errors with --read-inline-info=yes:
841 ==15377== Conditional jump or move depends on uninitialised value(s)
843 ==15377== at 0x80484EA: fun_d (inlinfo.c:6)
844 ==15377== by 0x80484EA: fun_c (inlinfo.c:14)
845 ==15377== by 0x80484EA: fun_b (inlinfo.c:20)
846 ==15377== by 0x80484EA: fun_a (inlinfo.c:26)
847 ==15377== by 0x80484EA: main (inlinfo.c:33)
848 ==15377==
849 ==15377== Conditional jump or move depends on uninitialised value(s)
850 ==15377== at 0x8048550: fun_d (inlinfo.c:6)
851 ==15377== by 0x8048550: fun_noninline (inlinfo.c:41)
852 ==15377== by 0x804850E: main (inlinfo.c:34)
853 ==15377==
854 ==15377== Conditional jump or move depends on uninitialised value(s)
855 ==15377== at 0x8048520: fun_d (inlinfo.c:6)
856 ==15377== by 0x8048520: main (inlinfo.c:35)
857 --read-var-info=<yes|no> [default: no]
859 When enabled, Valgrind will read information about variable types
860 and locations from DWARF3 debug info. This slows Valgrind startup
861 significantly and makes it use significantly more memory, but for
862 the tools that can take advantage of it (Memcheck, Helgrind, DRD)
863 it can result in more precise error messages. For example, here are
864 some standard errors issued by Memcheck:
865 ==15363== Uninitialised byte(s) found during client check request
867 ==15363== at 0x80484A9: croak (varinfo1.c:28)
868 ==15363== by 0x8048544: main (varinfo1.c:55)
869 ==15363== Address 0x80497f7 is 7 bytes inside data symbol "global_i2"
870 ==15363==
871 ==15363== Uninitialised byte(s) found during client check request
872 ==15363== at 0x80484A9: croak (varinfo1.c:28)
873 ==15363== by 0x8048550: main (varinfo1.c:56)
874 ==15363== Address 0xbea0d0cc is on thread 1's stack
875 ==15363== in frame #1, created by main (varinfo1.c:45)
876 And here are the same errors with --read-var-info=yes:
878 ==15370== Uninitialised byte(s) found during client check request
880 ==15370== at 0x80484A9: croak (varinfo1.c:28)
881 ==15370== by 0x8048544: main (varinfo1.c:55)
882 ==15370== Location 0x80497f7 is 0 bytes inside global_i2[7],
883 ==15370== a global variable declared at varinfo1.c:41
884 ==15370==
885 ==15370== Uninitialised byte(s) found during client check request
886 ==15370== at 0x80484A9: croak (varinfo1.c:28)
887 ==15370== by 0x8048550: main (varinfo1.c:56)
888 ==15370== Location 0xbeb4a0cc is 0 bytes inside local var "local"
889 ==15370== declared at varinfo1.c:46, in frame #1 of thread 1
890 --vgdb-poll=<number> [default: 5000]
892 As part of its main loop, the Valgrind scheduler will poll to check
893 if some activity (such as an external command or some input from a
894 gdb) has to be handled by gdbserver. This activity poll will be
895 done after having run the given number of basic blocks (or slightly
896 more than the given number of basic blocks). This poll is quite
897 cheap so the default value is set relatively low. You might further
898 decrease this value if vgdb cannot use ptrace system call to
899 interrupt Valgrind if all threads are (most of the time) blocked in
900 a system call.
901 --vgdb-shadow-registers=no|yes [default: no]
903 When activated, gdbserver will expose the Valgrind shadow registers
904 to GDB. With this, the value of the Valgrind shadow registers can
905 be examined or changed using GDB. Exposing shadow registers only
906 works with GDB version 7.1 or later.
907 --vgdb-prefix=<prefix> [default: /tmp/vgdb-pipe]
909 To communicate with gdb/vgdb, the Valgrind gdbserver creates 3
910 files (2 named FIFOs and a mmap shared memory file). The prefix
911 option controls the directory and prefix for the creation of these
912 files.
913 --run-libc-freeres=<yes|no> [default: yes]
915 This option is only relevant when running Valgrind on Linux.
916 The GNU C library (libc.so), which is used by all programs, may
918 allocate memory for its own uses. Usually it doesn't bother to free
919 that memory when the program ends—there would be no point, since
920 the Linux kernel reclaims all process resources when a process
921 exits anyway, so it would just slow things down.
922 The glibc authors realised that this behaviour causes leak
924 checkers, such as Valgrind, to falsely report leaks in glibc, when
925 a leak check is done at exit. In order to avoid this, they provided
926 a routine called __libc_freeres specifically to make glibc release
927 all memory it has allocated. Memcheck therefore tries to run
928 __libc_freeres at exit.
929 Unfortunately, in some very old versions of glibc, __libc_freeres
931 is sufficiently buggy to cause segmentation faults. This was
932 particularly noticeable on Red Hat 7.1. So this option is provided
933 in order to inhibit the run of __libc_freeres. If your program
934 seems to run fine on Valgrind, but segfaults at exit, you may find
935 that --run-libc-freeres=no fixes that, although at the cost of
936 possibly falsely reporting space leaks in libc.so.
937 --run-cxx-freeres=<yes|no> [default: yes]
939 This option is only relevant when running Valgrind on Linux or
940 Solaris C++ programs.
941 The GNU Standard C++ library (libstdc++.so), which is used by all
943 C++ programs compiled with g++, may allocate memory for its own
944 uses. Usually it doesn't bother to free that memory when the
945 program ends—there would be no point, since the kernel reclaims all
946 process resources when a process exits anyway, so it would just
947 slow things down.
948 The gcc authors realised that this behaviour causes leak checkers,
950 such as Valgrind, to falsely report leaks in libstdc++, when a leak
951 check is done at exit. In order to avoid this, they provided a
952 routine called __gnu_cxx::__freeres specifically to make libstdc++
953 release all memory it has allocated. Memcheck therefore tries to
954 run __gnu_cxx::__freeres at exit.
955 For the sake of flexibility and unforeseen problems with
957 __gnu_cxx::__freeres, option --run-cxx-freeres=no exists, although
958 at the cost of possibly falsely reporting space leaks in
959 libstdc++.so.
960 --sim-hints=hint1,hint2,...
962 Pass miscellaneous hints to Valgrind which slightly modify the
963 simulated behaviour in nonstandard or dangerous ways, possibly to
964 help the simulation of strange features. By default no hints are
965 enabled. Use with caution! Currently known hints are:
966 · lax-ioctls: Be very lax about ioctl handling; the only
968 assumption is that the size is correct. Doesn't require the
969 full buffer to be initialised when writing. Without this, using
970 some device drivers with a large number of strange ioctl
971 commands becomes very tiresome.
972 · fuse-compatible: Enable special handling for certain system
974 calls that may block in a FUSE file-system. This may be
975 necessary when running Valgrind on a multi-threaded program
976 that uses one thread to manage a FUSE file-system and another
977 thread to access that file-system.
978 · enable-outer: Enable some special magic needed when the program
980 being run is itself Valgrind.
981 · no-inner-prefix: Disable printing a prefix > in front of each
983 stdout or stderr output line in an inner Valgrind being run by
984 an outer Valgrind. This is useful when running Valgrind
985 regression tests in an outer/inner setup. Note that the prefix
986 > will always be printed in front of the inner debug logging
987 lines.
988 · no-nptl-pthread-stackcache: This hint is only relevant when
990 running Valgrind on Linux; it is ignored on Solaris and Mac OS
991 X.
992 The GNU glibc pthread library (libpthread.so), which is used by
994 pthread programs, maintains a cache of pthread stacks. When a
995 pthread terminates, the memory used for the pthread stack and
996 some thread local storage related data structure are not always
997 directly released. This memory is kept in a cache (up to a
998 certain size), and is re-used if a new thread is started.
999 This cache causes the helgrind tool to report some false
1001 positive race condition errors on this cached memory, as
1002 helgrind does not understand the internal glibc cache
1003 synchronisation primitives. So, when using helgrind, disabling
1004 the cache helps to avoid false positive race conditions, in
1005 particular when using thread local storage variables (e.g.
1006 variables using the __thread qualifier).
1007 When using the memcheck tool, disabling the cache ensures the
1009 memory used by glibc to handle __thread variables is directly
1010 released when a thread terminates.
1011 Note: Valgrind disables the cache using some internal knowledge
1013 of the glibc stack cache implementation and by examining the
1014 debug information of the pthread library. This technique is
1015 thus somewhat fragile and might not work for all glibc
1016 versions. This has been successfully tested with various glibc
1017 versions (e.g. 2.11, 2.16, 2.18) on various platforms.
1018 · lax-doors: (Solaris only) Be very lax about door syscall
1020 handling over unrecognised door file descriptors. Does not
1021 require that full buffer is initialised when writing. Without
1022 this, programs using libdoor(3LIB) functionality with
1023 completely proprietary semantics may report large number of
1024 false positives.
1025 · fallback-llsc: (MIPS and ARM64 only): Enables an alternative
1027 implementation of Load-Linked (LL) and Store-Conditional (SC)
1028 instructions. The standard implementation gives more correct
1029 behaviour, but can cause indefinite looping on certain
1030 processor implementations that are intolerant of extra memory
1031 references between LL and SC. So far this is known only to
1032 happen on Cavium 3 cores. You should not need to use this flag,
1033 since the relevant cores are detected at startup and the
1034 alternative implementation is automatically enabled if
1035 necessary. There is no equivalent anti-flag: you cannot
1036 force-disable the alternative implementation, if it is
1037 automatically enabled. The underlying problem exists because
1038 the "standard" implementation of LL and SC is done by copying
1039 through LL and SC instructions into the instrumented code.
1040 However, tools may insert extra instrumentation memory
1041 references in between the LL and SC instructions. These memory
1042 references are not present in the original uninstrumented code,
1043 and their presence in the instrumented code can cause the SC
1044 instructions to persistently fail, leading to indefinite
1045 looping in LL-SC blocks. The alternative implementation gives
1046 correct behaviour of LL and SC instructions between threads in
1047 a process, up to and including the ABA scenario. It also gives
1048 correct behaviour between a Valgrinded thread and a
1049 non-Valgrinded thread running in a different process, that
1050 communicate via shared memory, but only up to and including
1051 correct CAS behaviour -- in this case the ABA scenario may not
1052 be correctly handled.
1053 --fair-sched=<no|yes|try> [default: no]
1055 The --fair-sched option controls the locking mechanism used by
1056 Valgrind to serialise thread execution. The locking mechanism
1057 controls the way the threads are scheduled, and different settings
1058 give different trade-offs between fairness and performance. For
1059 more details about the Valgrind thread serialisation scheme and its
1060 impact on performance and thread scheduling, see Scheduling and
1061 Multi-Thread Performance.
1062 · The value --fair-sched=yes activates a fair scheduler. In
1064 short, if multiple threads are ready to run, the threads will
1065 be scheduled in a round robin fashion. This mechanism is not
1066 available on all platforms or Linux versions. If not available,
1067 using --fair-sched=yes will cause Valgrind to terminate with an
1068 error.
1069 You may find this setting improves overall responsiveness if
1071 you are running an interactive multithreaded program, for
1072 example a web browser, on Valgrind.
1073 · The value --fair-sched=try activates fair scheduling if
1075 available on the platform. Otherwise, it will automatically
1076 fall back to --fair-sched=no.
1077 · The value --fair-sched=no activates a scheduler which does not
1079 guarantee fairness between threads ready to run, but which in
1080 general gives the highest performance.
1081 --kernel-variant=variant1,variant2,...
1083 Handle system calls and ioctls arising from minor variants of the
1084 default kernel for this platform. This is useful for running on
1085 hacked kernels or with kernel modules which support nonstandard
1086 ioctls, for example. Use with caution. If you don't understand what
1087 this option does then you almost certainly don't need it. Currently
1088 known variants are:
1089 · bproc: support the sys_broc system call on x86. This is for
1091 running on BProc, which is a minor variant of standard Linux
1092 which is sometimes used for building clusters.
1093 · android-no-hw-tls: some versions of the Android emulator for
1095 ARM do not provide a hardware TLS (thread-local state)
1096 register, and Valgrind crashes at startup. Use this variant to
1097 select software support for TLS.
1098 · android-gpu-sgx5xx: use this to support handling of proprietary
1100 ioctls for the PowerVR SGX 5XX series of GPUs on Android
1101 devices. Failure to select this does not cause stability
1102 problems, but may cause Memcheck to report false errors after
1103 the program performs GPU-specific ioctls.
1104 · android-gpu-adreno3xx: similarly, use this to support handling
1106 of proprietary ioctls for the Qualcomm Adreno 3XX series of
1107 GPUs on Android devices.
1108 --merge-recursive-frames=<number> [default: 0]
1110 Some recursive algorithms, for example balanced binary tree
1111 implementations, create many different stack traces, each
1112 containing cycles of calls. A cycle is defined as two identical
1113 program counter values separated by zero or more other program
1114 counter values. Valgrind may then use a lot of memory to store all
1115 these stack traces. This is a poor use of memory considering that
1116 such stack traces contain repeated uninteresting recursive calls
1117 instead of more interesting information such as the function that
1118 has initiated the recursive call.
1119 The option --merge-recursive-frames=<number> instructs Valgrind to
1121 detect and merge recursive call cycles having a size of up to
1122 <number> frames. When such a cycle is detected, Valgrind records
1123 the cycle in the stack trace as a unique program counter.
1124 The value 0 (the default) causes no recursive call merging. A value
1126 of 1 will cause stack traces of simple recursive algorithms (for
1127 example, a factorial implementation) to be collapsed. A value of 2
1128 will usually be needed to collapse stack traces produced by
1129 recursive algorithms such as binary trees, quick sort, etc. Higher
1130 values might be needed for more complex recursive algorithms.
1131 Note: recursive calls are detected by analysis of program counter
1133 values. They are not detected by looking at function names.
1134 --num-transtab-sectors=<number> [default: 6 for Android platforms, 16
1136 for all others]
1137 Valgrind translates and instruments your program's machine code in
1138 small fragments (basic blocks). The translations are stored in a
1139 translation cache that is divided into a number of sections
1140 (sectors). If the cache is full, the sector containing the oldest
1141 translations is emptied and reused. If these old translations are
1142 needed again, Valgrind must re-translate and re-instrument the
1143 corresponding machine code, which is expensive. If the "executed
1144 instructions" working set of a program is big, increasing the
1145 number of sectors may improve performance by reducing the number of
1146 re-translations needed. Sectors are allocated on demand. Once
1147 allocated, a sector can never be freed, and occupies considerable
1148 space, depending on the tool and the value of
1149 --avg-transtab-entry-size (about 40 MB per sector for Memcheck).
1150 Use the option --stats=yes to obtain precise information about the
1151 memory used by a sector and the allocation and recycling of
1152 sectors.
1153 --avg-transtab-entry-size=<number> [default: 0, meaning use tool
1155 provided default]
1156 Average size of translated basic block. This average size is used
1157 to dimension the size of a sector. Each tool provides a default
1158 value to be used. If this default value is too small, the
1159 translation sectors will become full too quickly. If this default
1160 value is too big, a significant part of the translation sector
1161 memory will be unused. Note that the average size of a basic block
1162 translation depends on the tool, and might depend on tool options.
1163 For example, the memcheck option --track-origins=yes increases the
1164 size of the basic block translations. Use --avg-transtab-entry-size
1165 to tune the size of the sectors, either to gain memory or to avoid
1166 too many retranslations.
1167 --aspace-minaddr=<address> [default: depends on the platform]
1169 To avoid potential conflicts with some system libraries, Valgrind
1170 does not use the address space below --aspace-minaddr value,
1171 keeping it reserved in case a library specifically requests memory
1172 in this region. So, some "pessimistic" value is guessed by Valgrind
1173 depending on the platform. On linux, by default, Valgrind avoids
1174 using the first 64MB even if typically there is no conflict in this
1175 complete zone. You can use the option --aspace-minaddr to have your
1176 memory hungry application benefitting from more of this lower
1177 memory. On the other hand, if you encounter a conflict, increasing
1178 aspace-minaddr value might solve it. Conflicts will typically
1179 manifest themselves with mmap failures in the low range of the
1180 address space. The provided address must be page aligned and must
1181 be equal or bigger to 0x1000 (4KB). To find the default value on
1182 your platform, do something such as valgrind -d -d date 2>&1 | grep
1183 -i minaddr. Values lower than 0x10000 (64KB) are known to create
1184 problems on some distributions.
1185 --valgrind-stacksize=<number> [default: 1MB]
1187 For each thread, Valgrind needs its own 'private' stack. The
1188 default size for these stacks is largely dimensioned, and so should
1189 be sufficient in most cases. In case the size is too small,
1190 Valgrind will segfault. Before segfaulting, a warning might be
1191 produced by Valgrind when approaching the limit.
1192 Use the option --valgrind-stacksize if such an (unlikely) warning
1194 is produced, or Valgrind dies due to a segmentation violation. Such
1195 segmentation violations have been seen when demangling huge C++
1196 symbols.
1197 If your application uses many threads and needs a lot of memory,
1199 you can gain some memory by reducing the size of these Valgrind
1200 stacks using the option --valgrind-stacksize.
1201 --show-emwarns=<yes|no> [default: no]
1203 When enabled, Valgrind will emit warnings about its CPU emulation
1204 in certain cases. These are usually not interesting.
1205 --require-text-symbol=:sonamepatt:fnnamepatt
1207 When a shared object whose soname matches sonamepatt is loaded into
1208 the process, examine all the text symbols it exports. If none of
1209 those match fnnamepatt, print an error message and abandon the run.
1210 This makes it possible to ensure that the run does not continue
1211 unless a given shared object contains a particular function name.
1212 Both sonamepatt and fnnamepatt can be written using the usual ?
1214 and * wildcards. For example: ":*libc.so*:foo?bar". You may use
1215 characters other than a colon to separate the two patterns. It is
1216 only important that the first character and the separator character
1217 are the same. For example, the above example could also be written
1218 "Q*libc.so*Qfoo?bar". Multiple
1219 --require-text-symbol flags are allowed, in which case shared
1220 objects that are loaded into the process will be checked against
1221 all of them.
1222 The purpose of this is to support reliable usage of marked-up
1224 libraries. For example, suppose we have a version of GCC's
1225 libgomp.so which has been marked up with annotations to support
1226 Helgrind. It is only too easy and confusing to load the wrong,
1227 un-annotated libgomp.so into the application. So the idea is: add a
1228 text symbol in the marked-up library, for example
1229 annotated_for_helgrind_3_6, and then give the flag
1230 --require-text-symbol=:*libgomp*so*:annotated_for_helgrind_3_6 so
1231 that when libgomp.so is loaded, Valgrind scans its symbol table,
1232 and if the symbol isn't present the run is aborted, rather than
1233 continuing silently with the un-marked-up library. Note that you
1234 should put the entire flag in quotes to stop shells expanding up
1235 the * and ? wildcards.
1236 --soname-synonyms=syn1=pattern1,syn2=pattern2,...
1238 When a shared library is loaded, Valgrind checks for functions in
1239 the library that must be replaced or wrapped. For example, Memcheck
1240 replaces some string and memory functions (strchr, strlen, strcpy,
1241 memchr, memcpy, memmove, etc.) with its own versions. Such
1242 replacements are normally done only in shared libraries whose
1243 soname matches a predefined soname pattern (e.g. libc.so* on
1244 linux). By default, no replacement is done for a statically linked
1245 binary or for alternative libraries, except for the allocation
1246 functions (malloc, free, calloc, memalign, realloc, operator new,
1247 operator delete, etc.) Such allocation functions are intercepted by
1248 default in any shared library or in the executable if they are
1249 exported as global symbols. This means that if a replacement
1250 allocation library such as tcmalloc is found, its functions are
1251 also intercepted by default. In some cases, the replacements allow
1252 --soname-synonyms to specify one additional synonym pattern, giving
1253 flexibility in the replacement. Or to prevent interception of all
1254 public allocation symbols.
1255 Currently, this flexibility is only allowed for the malloc related
1257 functions, using the synonym somalloc. This synonym is usable for
1258 all tools doing standard replacement of malloc related functions
1259 (e.g. memcheck, massif, drd, helgrind, exp-dhat, exp-sgcheck).
1260 · Alternate malloc library: to replace the malloc related
1262 functions in a specific alternate library with soname
1263 mymalloclib.so (and not in any others), give the option
1264 --soname-synonyms=somalloc=mymalloclib.so. A pattern can be
1265 used to match multiple libraries sonames. For example,
1266 --soname-synonyms=somalloc=*tcmalloc* will match the soname of
1267 all variants of the tcmalloc library (native, debug, profiled,
1268 ... tcmalloc variants).
1269 Note: the soname of a elf shared library can be retrieved using
1271 the readelf utility.
1272 · Replacements in a statically linked library are done by using
1274 the NONE pattern. For example, if you link with libtcmalloc.a,
1275 and only want to intercept the malloc related functions in the
1276 executable (and standard libraries) themselves, but not any
1277 other shared libraries, you can give the option
1278 --soname-synonyms=somalloc=NONE. Note that a NONE pattern will
1279 match the main executable and any shared library having no
1280 soname.
1281 · To run a "default" Firefox build for Linux, in which JEMalloc
1283 is linked in to the main executable, use
1284 --soname-synonyms=somalloc=NONE.
1285 · To only intercept allocation symbols in the default system
1287 libraries, but not in any other shared library or the
1288 executable defining public malloc or operator new related
1289 functions use a non-existing library name like
1290 --soname-synonyms=somalloc=nouserintercepts (where
1291 nouserintercepts can be any non-existing library name).
1292 · Shared library of the dynamic (runtime) linker is excluded from
1294 searching for global public symbols, such as those for the
1295 malloc related functions (identified by somalloc synonym).
1296 --progress-interval=<number> [default: 0, meaning 'disabled']
1298 This is an enhancement to Valgrind's debugging output. It is
1299 unlikely to be of interest to end users.
1300 When number is set to a non-zero value, Valgrind will print a
1302 one-line progress summary every number seconds. Valid settings for
1303 number are between 0 and 3600 inclusive. Here's some example output
1304 with number set to 10:
1305 PROGRESS: U 110s, W 113s, 97.3% CPU, EvC 414.79M, TIn 616.7k, TOut 0.5k, #thr 67
1307 PROGRESS: U 120s, W 124s, 96.8% CPU, EvC 505.27M, TIn 636.6k, TOut 3.0k, #thr 64
1308 PROGRESS: U 130s, W 134s, 97.0% CPU, EvC 574.90M, TIn 657.5k, TOut 3.0k, #thr 63
1309 Each line shows:
1311 · U: total user time.RE
1313 · W: total wallclock time.RE
1315 · CPU: overall average cpu use.RE
1317 · EvC: number of event checks. An event
1319 check is a backwards branch in the
1320 simulated program, so this is a
1321 measure of forward progress of the
1322 program.RE
1323 · TIn: number of code blocks instrumented
1325 by the JIT.RE
1326 · TOut: number of instrumented code
1328 blocks that have been thrown
1329 away.RE
1330 · #thr: number of threads in the
1332 program.RE
1333 From the progress of these, it is
1335 possible to observe:
1336 · when the program is compute bound
1338 (TIn
1339 rises slowly, EvC rises
1340 rapidly).RE
1341 · when the program is in a
1343 spinloop
1344 (TIn/TOut
1345 fixed, EvC rises
1346 rapidly).RE
1347 · when the program is
1349 JIT-bound (TIn
1350 rises rapidly).RE
1351 · when the program is
1353 rapidly discarding code
1354 (TOut rises
1355 rapidly).RE
1356 · when the program is
1358 about to achieve
1359 some expected state
1360 (EvC
1361 arrives at some
1362 value you
1363 expect).RE
1364 · when the
1366 program is
1367 idling (U rises
1368 more
1369 slowly than
1370 W).RE
1371
1374 There are also some options for debugging Valgrind itself. You
1375 shouldn't need to use them in the normal run of things. If you wish to
1376 see the list, use the --help-debug option.
1377
1379 --leak-check=<no|summary|yes|full> [default: summary]
1380 When enabled, search for memory leaks when the client program
1381 finishes. If set to summary, it says how many leaks occurred. If
1382 set to full or yes, each individual leak will be shown in detail
1383 and/or counted as an error, as specified by the options
1384 --show-leak-kinds and --errors-for-leak-kinds.
1385 --leak-resolution=<low|med|high> [default: high]
1387 When doing leak checking, determines how willing Memcheck is to
1388 consider different backtraces to be the same for the purposes of
1389 merging multiple leaks into a single leak report. When set to low,
1390 only the first two entries need match. When med, four entries have
1391 to match. When high, all entries need to match.
1392 For hardcore leak debugging, you probably want to use
1394 --leak-resolution=high together with --num-callers=40 or some such
1395 large number.
1396 Note that the --leak-resolution setting does not affect Memcheck's
1398 ability to find leaks. It only changes how the results are
1399 presented.
1400 --show-leak-kinds=<set> [default: definite,possible]
1402 Specifies the leak kinds to show in a full leak search, in one of
1403 the following ways:
1404 · a comma separated list of one or more of definite indirect
1406 possible reachable.
1407 · all to specify the complete set (all leak kinds). It is
1409 equivalent to
1410 --show-leak-kinds=definite,indirect,possible,reachable.
1411 · none for the empty set.
1413 --errors-for-leak-kinds=<set> [default: definite,possible]
1415 Specifies the leak kinds to count as errors in a full leak search.
1416 The <set> is specified similarly to --show-leak-kinds
1417 --leak-check-heuristics=<set> [default: all]
1419 Specifies the set of leak check heuristics to be used during leak
1420 searches. The heuristics control which interior pointers to a block
1421 cause it to be considered as reachable. The heuristic set is
1422 specified in one of the following ways:
1423 · a comma separated list of one or more of stdstring length64
1425 newarray multipleinheritance.
1426 · all to activate the complete set of heuristics. It is
1428 equivalent to
1429 --leak-check-heuristics=stdstring,length64,newarray,multipleinheritance.
1430 · none for the empty set.
1432 Note that these heuristics are dependent on the layout of the
1434 objects produced by the C++ compiler. They have been tested with
1435 some gcc versions (e.g. 4.4 and 4.7). They might not work properly
1436 with other C++ compilers.
1437 --show-reachable=<yes|no> , --show-possibly-lost=<yes|no>
1439 These options provide an alternative way to specify the leak kinds
1440 to show:
1441 · --show-reachable=no --show-possibly-lost=yes is equivalent to
1443 --show-leak-kinds=definite,possible.
1444 · --show-reachable=no --show-possibly-lost=no is equivalent to
1446 --show-leak-kinds=definite.
1447 · --show-reachable=yes is equivalent to --show-leak-kinds=all.
1449 Note that --show-possibly-lost=no has no effect if
1451 --show-reachable=yes is specified.
1452 --xtree-leak=<no|yes> [no]
1454 If set to yes, the results for the leak search done at exit will be
1455 output in a 'Callgrind Format' execution tree file. Note that this
1456 automatically sets the option --leak-check=full. The produced file
1457 will contain the following events:
1458 · RB : Reachable Bytes
1460 · PB : Possibly lost Bytes
1462 · IB : Indirectly lost Bytes
1464 · DB : Definitely lost Bytes (direct plus indirect)
1466 · DIB : Definitely Indirectly lost Bytes (subset of DB)
1468 · RBk : reachable Blocks
1470 · PBk : Possibly lost Blocks
1472 · IBk : Indirectly lost Blocks
1474 · DBk : Definitely lost Blocks
1476 The increase or decrease for all events above will also be output
1478 in the file to provide the delta (increase or decrease) between 2
1479 successive leak searches. For example, iRB is the increase of the
1480 RB event, dPBk is the decrease of PBk event. The values for the
1481 increase and decrease events will be zero for the first leak search
1482 done.
1483 See ??? for a detailed explanation about execution trees.
1485 --xtree-leak-file=<filename> [default: xtleak.kcg.%p]
1487 Specifies that Valgrind should produce the xtree leak report in the
1488 specified file. Any %p, %q or %n sequences appearing in the
1489 filename are expanded in exactly the same way as they are for
1490 --log-file. See the description of --log-file for details.
1491 See ??? for a detailed explanation about execution trees formats.
1493 --undef-value-errors=<yes|no> [default: yes]
1495 Controls whether Memcheck reports uses of undefined value errors.
1496 Set this to no if you don't want to see undefined value errors. It
1497 also has the side effect of speeding up Memcheck somewhat.
1498 AddrCheck (removed in Valgrind 3.1.0) functioned like Memcheck with
1499 --undef-value-errors=no.
1500 --track-origins=<yes|no> [default: no]
1502 Controls whether Memcheck tracks the origin of uninitialised
1503 values. By default, it does not, which means that although it can
1504 tell you that an uninitialised value is being used in a dangerous
1505 way, it cannot tell you where the uninitialised value came from.
1506 This often makes it difficult to track down the root problem.
1507 When set to yes, Memcheck keeps track of the origins of all
1509 uninitialised values. Then, when an uninitialised value error is
1510 reported, Memcheck will try to show the origin of the value. An
1511 origin can be one of the following four places: a heap block, a
1512 stack allocation, a client request, or miscellaneous other sources
1513 (eg, a call to brk).
1514 For uninitialised values originating from a heap block, Memcheck
1516 shows where the block was allocated. For uninitialised values
1517 originating from a stack allocation, Memcheck can tell you which
1518 function allocated the value, but no more than that -- typically it
1519 shows you the source location of the opening brace of the function.
1520 So you should carefully check that all of the function's local
1521 variables are initialised properly.
1522 Performance overhead: origin tracking is expensive. It halves
1524 Memcheck's speed and increases memory use by a minimum of 100MB,
1525 and possibly more. Nevertheless it can drastically reduce the
1526 effort required to identify the root cause of uninitialised value
1527 errors, and so is often a programmer productivity win, despite
1528 running more slowly.
1529 Accuracy: Memcheck tracks origins quite accurately. To avoid very
1531 large space and time overheads, some approximations are made. It is
1532 possible, although unlikely, that Memcheck will report an incorrect
1533 origin, or not be able to identify any origin.
1534 Note that the combination --track-origins=yes and
1536 --undef-value-errors=no is nonsensical. Memcheck checks for and
1537 rejects this combination at startup.
1538 --partial-loads-ok=<yes|no> [default: yes]
1540 Controls how Memcheck handles 32-, 64-, 128- and 256-bit naturally
1541 aligned loads from addresses for which some bytes are addressable
1542 and others are not. When yes, such loads do not produce an address
1543 error. Instead, loaded bytes originating from illegal addresses are
1544 marked as uninitialised, and those corresponding to legal addresses
1545 are handled in the normal way.
1546 When no, loads from partially invalid addresses are treated the
1548 same as loads from completely invalid addresses: an illegal-address
1549 error is issued, and the resulting bytes are marked as initialised.
1550 Note that code that behaves in this way is in violation of the ISO
1552 C/C++ standards, and should be considered broken. If at all
1553 possible, such code should be fixed.
1554 --expensive-definedness-checks=<no|auto|yes> [default: auto]
1556 Controls whether Memcheck should employ more precise but also more
1557 expensive (time consuming) instrumentation when checking the
1558 definedness of certain values. In particular, this affects the
1559 instrumentation of integer adds, subtracts and equality
1560 comparisons.
1561 Selecting --expensive-definedness-checks=yes causes Memcheck to use
1563 the most accurate analysis possible. This minimises false error
1564 rates but can cause up to 30% performance degradation.
1565 Selecting --expensive-definedness-checks=no causes Memcheck to use
1567 the cheapest instrumentation possible. This maximises performance
1568 but will normally give an unusably high false error rate.
1569 The default setting, --expensive-definedness-checks=auto, is
1571 strongly recommended. This causes Memcheck to use the minimum of
1572 expensive instrumentation needed to achieve the same false error
1573 rate as --expensive-definedness-checks=yes. It also enables an
1574 instrumentation-time analysis pass which aims to further reduce the
1575 costs of accurate instrumentation. Overall, the performance loss is
1576 generally around 5% relative to --expensive-definedness-checks=no,
1577 although this is strongly workload dependent. Note that the exact
1578 instrumentation settings in this mode are architecture dependent.
1579 --keep-stacktraces=alloc|free|alloc-and-free|alloc-then-free|none
1581 [default: alloc-and-free]
1582 Controls which stack trace(s) to keep for malloc'd and/or free'd
1583 blocks.
1584 With alloc-then-free, a stack trace is recorded at allocation time,
1586 and is associated with the block. When the block is freed, a second
1587 stack trace is recorded, and this replaces the allocation stack
1588 trace. As a result, any "use after free" errors relating to this
1589 block can only show a stack trace for where the block was freed.
1590 With alloc-and-free, both allocation and the deallocation stack
1592 traces for the block are stored. Hence a "use after free" error
1593 will show both, which may make the error easier to diagnose.
1594 Compared to alloc-then-free, this setting slightly increases
1595 Valgrind's memory use as the block contains two references instead
1596 of one.
1597 With alloc, only the allocation stack trace is recorded (and
1599 reported). With free, only the deallocation stack trace is recorded
1600 (and reported). These values somewhat decrease Valgrind's memory
1601 and cpu usage. They can be useful depending on the error types you
1602 are searching for and the level of detail you need to analyse them.
1603 For example, if you are only interested in memory leak errors, it
1604 is sufficient to record the allocation stack traces.
1605 With none, no stack traces are recorded for malloc and free
1607 operations. If your program allocates a lot of blocks and/or
1608 allocates/frees from many different stack traces, this can
1609 significantly decrease cpu and/or memory required. Of course, few
1610 details will be reported for errors related to heap blocks.
1611 Note that once a stack trace is recorded, Valgrind keeps the stack
1613 trace in memory even if it is not referenced by any block. Some
1614 programs (for example, recursive algorithms) can generate a huge
1615 number of stack traces. If Valgrind uses too much memory in such
1616 circumstances, you can reduce the memory required with the options
1617 --keep-stacktraces and/or by using a smaller value for the option
1618 --num-callers.
1619 If you want to use --xtree-memory=full memory profiling (see ???
1621 ), then you cannot specify --keep-stacktraces=free or
1622 --keep-stacktraces=none.
1623 --freelist-vol=<number> [default: 20000000]
1625 When the client program releases memory using free (in C) or delete
1626 (C++), that memory is not immediately made available for
1627 re-allocation. Instead, it is marked inaccessible and placed in a
1628 queue of freed blocks. The purpose is to defer as long as possible
1629 the point at which freed-up memory comes back into circulation.
1630 This increases the chance that Memcheck will be able to detect
1631 invalid accesses to blocks for some significant period of time
1632 after they have been freed.
1633 This option specifies the maximum total size, in bytes, of the
1635 blocks in the queue. The default value is twenty million bytes.
1636 Increasing this increases the total amount of memory used by
1637 Memcheck but may detect invalid uses of freed blocks which would
1638 otherwise go undetected.
1639 --freelist-big-blocks=<number> [default: 1000000]
1641 When making blocks from the queue of freed blocks available for
1642 re-allocation, Memcheck will in priority re-circulate the blocks
1643 with a size greater or equal to --freelist-big-blocks. This ensures
1644 that freeing big blocks (in particular freeing blocks bigger than
1645 --freelist-vol) does not immediately lead to a re-circulation of
1646 all (or a lot of) the small blocks in the free list. In other
1647 words, this option increases the likelihood to discover dangling
1648 pointers for the "small" blocks, even when big blocks are freed.
1649 Setting a value of 0 means that all the blocks are re-circulated in
1651 a FIFO order.
1652 --workaround-gcc296-bugs=<yes|no> [default: no]
1654 When enabled, assume that reads and writes some small distance
1655 below the stack pointer are due to bugs in GCC 2.96, and does not
1656 report them. The "small distance" is 256 bytes by default. Note
1657 that GCC 2.96 is the default compiler on some ancient Linux
1658 distributions (RedHat 7.X) and so you may need to use this option.
1659 Do not use it if you do not have to, as it can cause real errors to
1660 be overlooked. A better alternative is to use a more recent GCC in
1661 which this bug is fixed.
1662 You may also need to use this option when working with GCC 3.X or
1664 4.X on 32-bit PowerPC Linux. This is because GCC generates code
1665 which occasionally accesses below the stack pointer, particularly
1666 for floating-point to/from integer conversions. This is in
1667 violation of the 32-bit PowerPC ELF specification, which makes no
1668 provision for locations below the stack pointer to be accessible.
1669 This option is deprecated as of version 3.12 and may be removed
1671 from future versions. You should instead use
1672 --ignore-range-below-sp to specify the exact range of offsets below
1673 the stack pointer that should be ignored. A suitable equivalent is
1674 --ignore-range-below-sp=1024-1.
1675 --ignore-range-below-sp=<number>-<number>
1677 This is a more general replacement for the deprecated
1678 --workaround-gcc296-bugs option. When specified, it causes Memcheck
1679 not to report errors for accesses at the specified offsets below
1680 the stack pointer. The two offsets must be positive decimal numbers
1681 and -- somewhat counterintuitively -- the first one must be larger,
1682 in order to imply a non-wraparound address range to ignore. For
1683 example, to ignore 4 byte accesses at 8192 bytes below the stack
1684 pointer, use --ignore-range-below-sp=8192-8189. Only one range may
1685 be specified.
1686 --show-mismatched-frees=<yes|no> [default: yes]
1688 When enabled, Memcheck checks that heap blocks are deallocated
1689 using a function that matches the allocating function. That is, it
1690 expects free to be used to deallocate blocks allocated by malloc,
1691 delete for blocks allocated by new, and delete[] for blocks
1692 allocated by new[]. If a mismatch is detected, an error is
1693 reported. This is in general important because in some
1694 environments, freeing with a non-matching function can cause
1695 crashes.
1696 There is however a scenario where such mismatches cannot be
1698 avoided. That is when the user provides implementations of
1699 new/new[] that call malloc and of delete/delete[] that call free,
1700 and these functions are asymmetrically inlined. For example,
1701 imagine that delete[] is inlined but new[] is not. The result is
1702 that Memcheck "sees" all delete[] calls as direct calls to free,
1703 even when the program source contains no mismatched calls.
1704 This causes a lot of confusing and irrelevant error reports.
1706 --show-mismatched-frees=no disables these checks. It is not
1707 generally advisable to disable them, though, because you may miss
1708 real errors as a result.
1709 --ignore-ranges=0xPP-0xQQ[,0xRR-0xSS]
1711 Any ranges listed in this option (and multiple ranges can be
1712 specified, separated by commas) will be ignored by Memcheck's
1713 addressability checking.
1714 --malloc-fill=<hexnumber>
1716 Fills blocks allocated by malloc, new, etc, but not by calloc, with
1717 the specified byte. This can be useful when trying to shake out
1718 obscure memory corruption problems. The allocated area is still
1719 regarded by Memcheck as undefined -- this option only affects its
1720 contents. Note that --malloc-fill does not affect a block of memory
1721 when it is used as argument to client requests
1722 VALGRIND_MEMPOOL_ALLOC or VALGRIND_MALLOCLIKE_BLOCK.
1723 --free-fill=<hexnumber>
1725 Fills blocks freed by free, delete, etc, with the specified byte
1726 value. This can be useful when trying to shake out obscure memory
1727 corruption problems. The freed area is still regarded by Memcheck
1728 as not valid for access -- this option only affects its contents.
1729 Note that --free-fill does not affect a block of memory when it is
1730 used as argument to client requests VALGRIND_MEMPOOL_FREE or
1731 VALGRIND_FREELIKE_BLOCK.
1732
1734 --I1=<size>,<associativity>,<line size>
1735 Specify the size, associativity and line size of the level 1
1736 instruction cache.
1737 --D1=<size>,<associativity>,<line size>
1739 Specify the size, associativity and line size of the level 1 data
1740 cache.
1741 --LL=<size>,<associativity>,<line size>
1743 Specify the size, associativity and line size of the last-level
1744 cache.
1745 --cache-sim=no|yes [yes]
1747 Enables or disables collection of cache access and miss counts.
1748 --branch-sim=no|yes [no]
1750 Enables or disables collection of branch instruction and
1751 misprediction counts. By default this is disabled as it slows
1752 Cachegrind down by approximately 25%. Note that you cannot specify
1753 --cache-sim=no and --branch-sim=no together, as that would leave
1754 Cachegrind with no information to collect.
1755 --cachegrind-out-file=<file>
1757 Write the profile data to file rather than to the default output
1758 file, cachegrind.out.<pid>. The %p and %q format specifiers can be
1759 used to embed the process ID and/or the contents of an environment
1760 variable in the name, as is the case for the core option --log-
1761 file.
1762
1764 --callgrind-out-file=<file>
1765 Write the profile data to file rather than to the default output
1766 file, callgrind.out.<pid>. The %p and %q format specifiers can be
1767 used to embed the process ID and/or the contents of an environment
1768 variable in the name, as is the case for the core option --log-
1769 file. When multiple dumps are made, the file name is modified
1770 further; see below.
1771 --dump-line=<no|yes> [default: yes]
1773 This specifies that event counting should be performed at source
1774 line granularity. This allows source annotation for sources which
1775 are compiled with debug information (-g).
1776 --dump-instr=<no|yes> [default: no]
1778 This specifies that event counting should be performed at
1779 per-instruction granularity. This allows for assembly code
1780 annotation. Currently the results can only be displayed by
1781 KCachegrind.
1782 --compress-strings=<no|yes> [default: yes]
1784 This option influences the output format of the profile data. It
1785 specifies whether strings (file and function names) should be
1786 identified by numbers. This shrinks the file, but makes it more
1787 difficult for humans to read (which is not recommended in any
1788 case).
1789 --compress-pos=<no|yes> [default: yes]
1791 This option influences the output format of the profile data. It
1792 specifies whether numerical positions are always specified as
1793 absolute values or are allowed to be relative to previous numbers.
1794 This shrinks the file size.
1795 --combine-dumps=<no|yes> [default: no]
1797 When enabled, when multiple profile data parts are to be generated
1798 these parts are appended to the same output file. Not recommended.
1799 --dump-every-bb=<count> [default: 0, never]
1801 Dump profile data every count basic blocks. Whether a dump is
1802 needed is only checked when Valgrind's internal scheduler is run.
1803 Therefore, the minimum setting useful is about 100000. The count is
1804 a 64-bit value to make long dump periods possible.
1805 --dump-before=<function>
1807 Dump when entering function.
1808 --zero-before=<function>
1810 Zero all costs when entering function.
1811 --dump-after=<function>
1813 Dump when leaving function.
1814 --instr-atstart=<yes|no> [default: yes]
1816 Specify if you want Callgrind to start simulation and profiling
1817 from the beginning of the program. When set to no, Callgrind will
1818 not be able to collect any information, including calls, but it
1819 will have at most a slowdown of around 4, which is the minimum
1820 Valgrind overhead. Instrumentation can be interactively enabled via
1821 callgrind_control -i on.
1822 Note that the resulting call graph will most probably not contain
1824 main, but will contain all the functions executed after
1825 instrumentation was enabled. Instrumentation can also
1826 programatically enabled/disabled. See the Callgrind include file
1827 callgrind.h for the macro you have to use in your source code.
1828 For cache simulation, results will be less accurate when switching
1830 on instrumentation later in the program run, as the simulator
1831 starts with an empty cache at that moment. Switch on event
1832 collection later to cope with this error.
1833 --collect-atstart=<yes|no> [default: yes]
1835 Specify whether event collection is enabled at beginning of the
1836 profile run.
1837 To only look at parts of your program, you have two possibilities:
1839 1. Zero event counters before entering the program part you want
1841 to profile, and dump the event counters to a file after leaving
1842 that program part.
1843 2. Switch on/off collection state as needed to only see event
1845 counters happening while inside of the program part you want to
1846 profile.
1847 The second option can be used if the program part you want to
1849 profile is called many times. Option 1, i.e. creating a lot of
1850 dumps is not practical here.
1851 Collection state can be toggled at entry and exit of a given
1853 function with the option --toggle-collect. If you use this option,
1854 collection state should be disabled at the beginning. Note that the
1855 specification of --toggle-collect implicitly sets
1856 --collect-state=no.
1857 Collection state can be toggled also by inserting the client
1859 request CALLGRIND_TOGGLE_COLLECT ; at the needed code positions.
1860 --toggle-collect=<function>
1862 Toggle collection on entry/exit of function.
1863 --collect-jumps=<no|yes> [default: no]
1865 This specifies whether information for (conditional) jumps should
1866 be collected. As above, callgrind_annotate currently is not able to
1867 show you the data. You have to use KCachegrind to get jump arrows
1868 in the annotated code.
1869 --collect-systime=<no|yes> [default: no]
1871 This specifies whether information for system call times should be
1872 collected.
1873 --collect-bus=<no|yes> [default: no]
1875 This specifies whether the number of global bus events executed
1876 should be collected. The event type "Ge" is used for these events.
1877 --cache-sim=<yes|no> [default: no]
1879 Specify if you want to do full cache simulation. By default, only
1880 instruction read accesses will be counted ("Ir"). With cache
1881 simulation, further event counters are enabled: Cache misses on
1882 instruction reads ("I1mr"/"ILmr"), data read accesses ("Dr") and
1883 related cache misses ("D1mr"/"DLmr"), data write accesses ("Dw")
1884 and related cache misses ("D1mw"/"DLmw"). For more information, see
1885 Cachegrind: a cache and branch-prediction profiler.
1886 --branch-sim=<yes|no> [default: no]
1888 Specify if you want to do branch prediction simulation. Further
1889 event counters are enabled: Number of executed conditional branches
1890 and related predictor misses ("Bc"/"Bcm"), executed indirect jumps
1891 and related misses of the jump address predictor ("Bi"/"Bim").
1892
1894 --free-is-write=no|yes [default: no]
1895 When enabled (not the default), Helgrind treats freeing of heap
1896 memory as if the memory was written immediately before the free.
1897 This exposes races where memory is referenced by one thread, and
1898 freed by another, but there is no observable synchronisation event
1899 to ensure that the reference happens before the free.
1900 This functionality is new in Valgrind 3.7.0, and is regarded as
1902 experimental. It is not enabled by default because its interaction
1903 with custom memory allocators is not well understood at present.
1904 User feedback is welcomed.
1905 --track-lockorders=no|yes [default: yes]
1907 When enabled (the default), Helgrind performs lock order
1908 consistency checking. For some buggy programs, the large number of
1909 lock order errors reported can become annoying, particularly if
1910 you're only interested in race errors. You may therefore find it
1911 helpful to disable lock order checking.
1912 --history-level=none|approx|full [default: full]
1914 --history-level=full (the default) causes Helgrind collects enough
1915 information about "old" accesses that it can produce two stack
1916 traces in a race report -- both the stack trace for the current
1917 access, and the trace for the older, conflicting access. To limit
1918 memory usage, "old" accesses stack traces are limited to a maximum
1919 of 8 entries, even if --num-callers value is bigger.
1920 Collecting such information is expensive in both speed and memory,
1922 particularly for programs that do many inter-thread synchronisation
1923 events (locks, unlocks, etc). Without such information, it is more
1924 difficult to track down the root causes of races. Nonetheless, you
1925 may not need it in situations where you just want to check for the
1926 presence or absence of races, for example, when doing regression
1927 testing of a previously race-free program.
1928 --history-level=none is the opposite extreme. It causes Helgrind
1930 not to collect any information about previous accesses. This can be
1931 dramatically faster than --history-level=full.
1932 --history-level=approx provides a compromise between these two
1934 extremes. It causes Helgrind to show a full trace for the later
1935 access, and approximate information regarding the earlier access.
1936 This approximate information consists of two stacks, and the
1937 earlier access is guaranteed to have occurred somewhere between
1938 program points denoted by the two stacks. This is not as useful as
1939 showing the exact stack for the previous access (as
1940 --history-level=full does), but it is better than nothing, and it
1941 is almost as fast as --history-level=none.
1942 --delta-stacktrace=no|yes [default: yes on linux amd64/x86]
1944 This flag only has any effect at --history-level=full.
1945 --delta-stacktrace configures the way Helgrind captures the
1947 stacktraces for the option --history-level=full. Such a stacktrace
1948 is typically needed each time a new piece of memory is read or
1949 written in a basic block of instructions.
1950 --delta-stacktrace=no causes Helgrind to compute a full history
1952 stacktrace from the unwind info each time a stacktrace is needed.
1953 --delta-stacktrace=yes indicates to Helgrind to derive a new
1955 stacktrace from the previous stacktrace, as long as there was no
1956 call instruction, no return instruction, or any other instruction
1957 changing the call stack since the previous stacktrace was captured.
1958 If no such instruction was executed, the new stacktrace can be
1959 derived from the previous stacktrace by just changing the top frame
1960 to the current program counter. This option can speed up Helgrind
1961 by 25% when using --history-level=full.
1962 The following aspects have to be considered when using
1964 --delta-stacktrace=yes :
1965 · In some cases (for example in a function prologue), the
1967 valgrind unwinder might not properly unwind the
1968 stack, due to some
1969 limitations and/or due to wrong unwind info. When
1970 using
1971 --delta-stacktrace=yes, the wrong stack trace
1972 captured in the
1973 function prologue will be kept till the next call
1974 or return.
1975 .RE
1976 · On the other hand, --delta-stacktrace=yes sometimes helps
1978 to
1979 obtain a correct stacktrace, for example when
1980 the unwind info allows
1981 a correct stacktrace to be done in the
1982 beginning of the sequence,
1983 but not later on in the instruction
1984 sequence..RE
1985 · Determining which instructions are changing the
1987 callstack is
1988 partially based on platform dependent
1989 heuristics, which have to be
1990 tuned/validated specifically for the
1991 platform. Also, unwinding in a
1992 function prologue must be good enough to
1993 allow using
1994 --delta-stacktrace=yes. Currently, the
1995 option --delta-stacktrace=yes
1996 has been reasonably validated only on linux
1997 x86 32 bits and linux
1998 amd64 64 bits. For more details about how
1999 to validate
2000 --delta-stacktrace=yes, see debug option
2001 --hg-sanity-flags and the
2002 function check_cached_rcec_ok in
2003 libhb_core.c..RE
2004 --conflict-cache-size=N [default: 1000000]
2007 This flag only has any effect at --history-level=full.
2008 Information about "old" conflicting accesses is stored
2010 in a cache of limited size, with LRU-style management.
2011 This is necessary because it isn't practical to store a
2012 stack trace for every single memory access made by the
2013 program. Historical information on not recently
2014 accessed locations is periodically discarded, to free
2015 up space in the cache.
2016 This option controls the size of the cache, in terms of
2018 the number of different memory addresses for which
2019 conflicting access information is stored. If you find
2020 that Helgrind is showing race errors with only one
2021 stack instead of the expected two stacks, try
2022 increasing this value.
2023 The minimum value is 10,000 and the maximum is
2025 30,000,000 (thirty times the default value). Increasing
2026 the value by 1 increases Helgrind's memory requirement
2027 by very roughly 100 bytes, so the maximum value will
2028 easily eat up three extra gigabytes or so of memory.
2029 --check-stack-refs=no|yes [default: yes]
2031 By default Helgrind checks all data memory accesses
2032 made by your program. This flag enables you to skip
2033 checking for accesses to thread stacks (local
2034 variables). This can improve performance, but comes at
2035 the cost of missing races on stack-allocated data.
2036 --ignore-thread-creation=<yes|no> [default: no]
2038 Controls whether all activities during thread creation
2039 should be ignored. By default enabled only on Solaris.
2040 Solaris provides higher throughput, parallelism and
2041 scalability than other operating systems, at the cost
2042 of more fine-grained locking activity. This means for
2043 example that when a thread is created under glibc, just
2044 one big lock is used for all thread setup. Solaris libc
2045 uses several fine-grained locks and the creator thread
2046 resumes its activities as soon as possible, leaving for
2047 example stack and TLS setup sequence to the created
2048 thread. This situation confuses Helgrind as it assumes
2049 there is some false ordering in place between creator
2050 and created thread; and therefore many types of race
2051 conditions in the application would not be reported. To
2052 prevent such false ordering, this command line option
2053 is set to yes by default on Solaris. All activity
2054 (loads, stores, client requests) is therefore ignored
2055 during:
2056 · pthread_create() call in the creator thread
2058 · thread creation phase (stack and TLS setup) in the
2060 created thread
2061 Also new memory allocated during thread creation is
2063 untracked, that is race reporting is suppressed there.
2064 DRD does the same thing implicitly. This is necessary
2065 because Solaris libc caches many objects and reuses
2066 them for different threads and that confuses Helgrind.
2067
2069 --check-stack-var=<yes|no> [default: no]
2070 Controls whether DRD detects data races on stack variables.
2071 Verifying stack variables is disabled by default because most
2072 programs do not share stack variables over threads.
2073 --exclusive-threshold=<n> [default: off]
2075 Print an error message if any mutex or writer lock has been held
2076 longer than the time specified in milliseconds. This option enables
2077 the detection of lock contention.
2078 --join-list-vol=<n> [default: 10]
2080 Data races that occur between a statement at the end of one thread
2081 and another thread can be missed if memory access information is
2082 discarded immediately after a thread has been joined. This option
2083 allows to specify for how many joined threads memory access
2084 information should be retained.
2085 --first-race-only=<yes|no> [default: no]
2087 Whether to report only the first data race that has been detected
2088 on a memory location or all data races that have been detected on a
2089 memory location.
2090 --free-is-write=<yes|no> [default: no]
2092 Whether to report races between accessing memory and freeing
2093 memory. Enabling this option may cause DRD to run slightly slower.
2094 Notes:
2095 · Don't enable this option when using custom memory allocators
2097 that use the VG_USERREQ__MALLOCLIKE_BLOCK and
2098 VG_USERREQ__FREELIKE_BLOCK because that would result in false
2099 positives.
2100 · Don't enable this option when using reference-counted objects
2102 because that will result in false positives, even when that
2103 code has been annotated properly with ANNOTATE_HAPPENS_BEFORE
2104 and ANNOTATE_HAPPENS_AFTER. See e.g. the output of the
2105 following command for an example: valgrind --tool=drd
2106 --free-is-write=yes drd/tests/annotate_smart_pointer.
2107 --report-signal-unlocked=<yes|no> [default: yes]
2109 Whether to report calls to pthread_cond_signal and
2110 pthread_cond_broadcast where the mutex associated with the signal
2111 through pthread_cond_wait or pthread_cond_timed_waitis not locked
2112 at the time the signal is sent. Sending a signal without holding a
2113 lock on the associated mutex is a common programming error which
2114 can cause subtle race conditions and unpredictable behavior. There
2115 exist some uncommon synchronization patterns however where it is
2116 safe to send a signal without holding a lock on the associated
2117 mutex.
2118 --segment-merging=<yes|no> [default: yes]
2120 Controls segment merging. Segment merging is an algorithm to limit
2121 memory usage of the data race detection algorithm. Disabling
2122 segment merging may improve the accuracy of the so-called 'other
2123 segments' displayed in race reports but can also trigger an out of
2124 memory error.
2125 --segment-merging-interval=<n> [default: 10]
2127 Perform segment merging only after the specified number of new
2128 segments have been created. This is an advanced configuration
2129 option that allows to choose whether to minimize DRD's memory usage
2130 by choosing a low value or to let DRD run faster by choosing a
2131 slightly higher value. The optimal value for this parameter depends
2132 on the program being analyzed. The default value works well for
2133 most programs.
2134 --shared-threshold=<n> [default: off]
2136 Print an error message if a reader lock has been held longer than
2137 the specified time (in milliseconds). This option enables the
2138 detection of lock contention.
2139 --show-confl-seg=<yes|no> [default: yes]
2141 Show conflicting segments in race reports. Since this information
2142 can help to find the cause of a data race, this option is enabled
2143 by default. Disabling this option makes the output of DRD more
2144 compact.
2145 --show-stack-usage=<yes|no> [default: no]
2147 Print stack usage at thread exit time. When a program creates a
2148 large number of threads it becomes important to limit the amount of
2149 virtual memory allocated for thread stacks. This option makes it
2150 possible to observe how much stack memory has been used by each
2151 thread of the client program. Note: the DRD tool itself allocates
2152 some temporary data on the client thread stack. The space necessary
2153 for this temporary data must be allocated by the client program
2154 when it allocates stack memory, but is not included in stack usage
2155 reported by DRD.
2156 --ignore-thread-creation=<yes|no> [default: no]
2158 Controls whether all activities during thread creation should be
2159 ignored. By default enabled only on Solaris. Solaris provides
2160 higher throughput, parallelism and scalability than other operating
2161 systems, at the cost of more fine-grained locking activity. This
2162 means for example that when a thread is created under glibc, just
2163 one big lock is used for all thread setup. Solaris libc uses
2164 several fine-grained locks and the creator thread resumes its
2165 activities as soon as possible, leaving for example stack and TLS
2166 setup sequence to the created thread. This situation confuses DRD
2167 as it assumes there is some false ordering in place between creator
2168 and created thread; and therefore many types of race conditions in
2169 the application would not be reported. To prevent such false
2170 ordering, this command line option is set to yes by default on
2171 Solaris. All activity (loads, stores, client requests) is therefore
2172 ignored during:
2173 · pthread_create() call in the creator thread
2175 · thread creation phase (stack and TLS setup) in the created
2177 thread
2178 --trace-addr=<address> [default: none]
2180 Trace all load and store activity for the specified address. This
2181 option may be specified more than once.
2182 --ptrace-addr=<address> [default: none]
2184 Trace all load and store activity for the specified address and
2185 keep doing that even after the memory at that address has been
2186 freed and reallocated.
2187 --trace-alloc=<yes|no> [default: no]
2189 Trace all memory allocations and deallocations. May produce a huge
2190 amount of output.
2191 --trace-barrier=<yes|no> [default: no]
2193 Trace all barrier activity.
2194 --trace-cond=<yes|no> [default: no]
2196 Trace all condition variable activity.
2197 --trace-fork-join=<yes|no> [default: no]
2199 Trace all thread creation and all thread termination events.
2200 --trace-hb=<yes|no> [default: no]
2202 Trace execution of the ANNOTATE_HAPPENS_BEFORE(),
2203 ANNOTATE_HAPPENS_AFTER() and ANNOTATE_HAPPENS_DONE() client
2204 requests.
2205 --trace-mutex=<yes|no> [default: no]
2207 Trace all mutex activity.
2208 --trace-rwlock=<yes|no> [default: no]
2210 Trace all reader-writer lock activity.
2211 --trace-semaphore=<yes|no> [default: no]
2213 Trace all semaphore activity.
2214
2216 --heap=<yes|no> [default: yes]
2217 Specifies whether heap profiling should be done.
2218 --heap-admin=<size> [default: 8]
2220 If heap profiling is enabled, gives the number of administrative
2221 bytes per block to use. This should be an estimate of the average,
2222 since it may vary. For example, the allocator used by glibc on
2223 Linux requires somewhere between 4 to 15 bytes per block, depending
2224 on various factors. That allocator also requires admin space for
2225 freed blocks, but Massif cannot account for this.
2226 --stacks=<yes|no> [default: no]
2228 Specifies whether stack profiling should be done. This option slows
2229 Massif down greatly, and so is off by default. Note that Massif
2230 assumes that the main stack has size zero at start-up. This is not
2231 true, but doing otherwise accurately is difficult. Furthermore,
2232 starting at zero better indicates the size of the part of the main
2233 stack that a user program actually has control over.
2234 --pages-as-heap=<yes|no> [default: no]
2236 Tells Massif to profile memory at the page level rather than at the
2237 malloc'd block level. See above for details.
2238 --depth=<number> [default: 30]
2240 Maximum depth of the allocation trees recorded for detailed
2241 snapshots. Increasing it will make Massif run somewhat more slowly,
2242 use more memory, and produce bigger output files.
2243 --alloc-fn=<name>
2245 Functions specified with this option will be treated as though they
2246 were a heap allocation function such as malloc. This is useful for
2247 functions that are wrappers to malloc or new, which can fill up the
2248 allocation trees with uninteresting information. This option can be
2249 specified multiple times on the command line, to name multiple
2250 functions.
2251 Note that the named function will only be treated this way if it is
2253 the top entry in a stack trace, or just below another function
2254 treated this way. For example, if you have a function malloc1 that
2255 wraps malloc, and malloc2 that wraps malloc1, just specifying
2256 --alloc-fn=malloc2 will have no effect. You need to specify
2257 --alloc-fn=malloc1 as well. This is a little inconvenient, but the
2258 reason is that checking for allocation functions is slow, and it
2259 saves a lot of time if Massif can stop looking through the stack
2260 trace entries as soon as it finds one that doesn't match rather
2261 than having to continue through all the entries.
2262 Note that C++ names are demangled. Note also that overloaded C++
2264 names must be written in full. Single quotes may be necessary to
2265 prevent the shell from breaking them up. For example:
2266 --alloc-fn='operator new(unsigned, std::nothrow_t const&)'
2268 --ignore-fn=<name>
2271 Any direct heap allocation (i.e. a call to malloc, new, etc, or a
2272 call to a function named by an --alloc-fn option) that occurs in a
2273 function specified by this option will be ignored. This is mostly
2274 useful for testing purposes. This option can be specified multiple
2275 times on the command line, to name multiple functions.
2276 Any realloc of an ignored block will also be ignored, even if the
2278 realloc call does not occur in an ignored function. This avoids the
2279 possibility of negative heap sizes if ignored blocks are shrunk
2280 with realloc.
2281 The rules for writing C++ function names are the same as for
2283 --alloc-fn above.
2284 --threshold=<m.n> [default: 1.0]
2286 The significance threshold for heap allocations, as a percentage of
2287 total memory size. Allocation tree entries that account for less
2288 than this will be aggregated. Note that this should be specified in
2289 tandem with ms_print's option of the same name.
2290 --peak-inaccuracy=<m.n> [default: 1.0]
2292 Massif does not necessarily record the actual global memory
2293 allocation peak; by default it records a peak only when the global
2294 memory allocation size exceeds the previous peak by at least 1.0%.
2295 This is because there can be many local allocation peaks along the
2296 way, and doing a detailed snapshot for every one would be expensive
2297 and wasteful, as all but one of them will be later discarded. This
2298 inaccuracy can be changed (even to 0.0%) via this option, but
2299 Massif will run drastically slower as the number approaches zero.
2300 --time-unit=<i|ms|B> [default: i]
2302 The time unit used for the profiling. There are three
2303 possibilities: instructions executed (i), which is good for most
2304 cases; real (wallclock) time (ms, i.e. milliseconds), which is
2305 sometimes useful; and bytes allocated/deallocated on the heap
2306 and/or stack (B), which is useful for very short-run programs, and
2307 for testing purposes, because it is the most reproducible across
2308 different machines.
2309 --detailed-freq=<n> [default: 10]
2311 Frequency of detailed snapshots. With --detailed-freq=1, every
2312 snapshot is detailed.
2313 --max-snapshots=<n> [default: 100]
2315 The maximum number of snapshots recorded. If set to N, for all
2316 programs except very short-running ones, the final number of
2317 snapshots will be between N/2 and N.
2318 --massif-out-file=<file> [default: massif.out.%p]
2320 Write the profile data to file rather than to the default output
2321 file, massif.out.<pid>. The %p and %q format specifiers can be used
2322 to embed the process ID and/or the contents of an environment
2323 variable in the name, as is the case for the core option --log-
2324 file.
2325
2327 There are no SGCheck-specific command-line options at present.
2328
2330 --bb-out-file=<name> [default: bb.out.%p]
2331 This option selects the name of the basic block vector file. The %p
2332 and %q format specifiers can be used to embed the process ID and/or
2333 the contents of an environment variable in the name, as is the case
2334 for the core option --log-file.
2335 --pc-out-file=<name> [default: pc.out.%p]
2337 This option selects the name of the PC file. This file holds
2338 program counter addresses and function name info for the various
2339 basic blocks. This can be used in conjunction with the basic block
2340 vector file to fast-forward via function names instead of just
2341 instruction counts. The %p and %q format specifiers can be used to
2342 embed the process ID and/or the contents of an environment variable
2343 in the name, as is the case for the core option --log-file.
2344 --interval-size=<number> [default: 100000000]
2346 This option selects the size of the interval to use. The default is
2347 100 million instructions, which is a commonly used value. Other
2348 sizes can be used; smaller intervals can help programs with
2349 finer-grained phases. However smaller interval size can lead to
2350 accuracy issues due to warm-up effects (When fast-forwarding the
2351 various architectural features will be un-initialized, and it will
2352 take some number of instructions before they "warm up" to the state
2353 a full simulation would be at without the fast-forwarding. Large
2354 interval sizes tend to mitigate this.)
2355 --instr-count-only [default: no]
2357 This option tells the tool to only display instruction count
2358 totals, and to not generate the actual basic block vector file.
2359 This is useful for debugging, and for gathering instruction count
2360 info without generating the large basic block vector files.
2361
2363 --basic-counts=<no|yes> [default: yes]
2364 When enabled, Lackey prints the following statistics and
2365 information about the execution of the client program:
2366 1. The number of calls to the function specified by the --fnname
2368 option (the default is main). If the program has had its
2369 symbols stripped, the count will always be zero.
2370 2. The number of conditional branches encountered and the number
2372 and proportion of those taken.
2373 3. The number of superblocks entered and completed by the program.
2375 Note that due to optimisations done by the JIT, this is not at
2376 all an accurate value.
2377 4. The number of guest (x86, amd64, ppc, etc.) instructions and IR
2379 statements executed. IR is Valgrind's RISC-like intermediate
2380 representation via which all instrumentation is done.
2381 5. Ratios between some of these counts.
2383 6. The exit code of the client program.
2385 --detailed-counts=<no|yes> [default: no]
2387 When enabled, Lackey prints a table containing counts of loads,
2388 stores and ALU operations, differentiated by their IR types. The IR
2389 types are identified by their IR name ("I1", "I8", ... "I128",
2390 "F32", "F64", and "V128").
2391 --trace-mem=<no|yes> [default: no]
2393 When enabled, Lackey prints the size and address of almost every
2394 memory access made by the program. See the comments at the top of
2395 the file lackey/lk_main.c for details about the output format, how
2396 it works, and inaccuracies in the address trace. Note that this
2397 option produces immense amounts of output.
2398 --trace-superblocks=<no|yes> [default: no]
2400 When enabled, Lackey prints out the address of every superblock (a
2401 single entry, multiple exit, linear chunk of code) executed by the
2402 program. This is primarily of interest to Valgrind developers. See
2403 the comments at the top of the file lackey/lk_main.c for details
2404 about the output format. Note that this option produces large
2405 amounts of output.
2406 --fnname=<name> [default: main]
2408 Changes the function for which calls are counted when
2409 --basic-counts=yes is specified.
2410
2412 cg_annotate(1), callgrind_annotate(1), callgrind_control(1),
2413 ms_print(1), $INSTALL/share/doc/valgrind/html/index.html or
2414 http://www.valgrind.org/docs/manual/index.html, Debugging your program
2415 using Valgrind's gdbserver and GDB[1] vgdb[2], Valgrind monitor
2416 commands[3], The Commentary[4], Scheduling and Multi-Thread
2417 Performance[5], Cachegrind: a cache and branch-prediction profiler[6].
2418
2420 See the AUTHORS file in the valgrind distribution for a comprehensive
2421 list of authors.
2422 This manpage was written by Andres Roldan <aroldan@debian.org> and the
2424 Valgrind developers.
2425
2427 1. Debugging your program using Valgrind's gdbserver and GDB
2428 http://www.valgrind.org/docs/manual/manual-core-adv.html#manual-core-adv.gdbserver
2429 2. vgdb
2431 http://www.valgrind.org/docs/manual/manual-core-adv.html#manual-core-adv.vgdb
2432 3. Valgrind monitor commands
2434 http://www.valgrind.org/docs/manual/manual-core-adv.html#manual-core-adv.valgrind-monitor-commands
2435 4. The Commentary
2437 http://www.valgrind.org/docs/manual/manual-core.html#manual-core.comment
2438 5. Scheduling and Multi-Thread Performance
2440 http://www.valgrind.org/docs/manual/manual-core.html#manual-core.pthreads_perf_sched
2441 6. Cachegrind: a cache and branch-prediction profiler
2443 http://www.valgrind.org/docs/manual/cg-manual.html
2444Release 3.14.0 10/09/2018 VALGRIND(1)