1 .\" Copyright (c) 2003-2008 Joseph Koshy
2 .\" Copyright (c) 2007 The FreeBSD Foundation
3 .\" All rights reserved.
5 .\" Redistribution and use in source and binary forms, with or without
6 .\" modification, are permitted provided that the following conditions
8 .\" 1. Redistributions of source code must retain the above copyright
9 .\" notice, this list of conditions and the following disclaimer.
10 .\" 2. Redistributions in binary form must reproduce the above copyright
11 .\" notice, this list of conditions and the following disclaimer in the
12 .\" documentation and/or other materials provided with the distribution.
14 .\" This software is provided by Joseph Koshy ``as is'' and
15 .\" any express or implied warranties, including, but not limited to, the
16 .\" implied warranties of merchantability and fitness for a particular purpose
17 .\" are disclaimed. in no event shall Joseph Koshy be liable
18 .\" for any direct, indirect, incidental, special, exemplary, or consequential
19 .\" damages (including, but not limited to, procurement of substitute goods
20 .\" or services; loss of use, data, or profits; or business interruption)
21 .\" however caused and on any theory of liability, whether in contract, strict
22 .\" liability, or tort (including negligence or otherwise) arising in any way
23 .\" out of the use of this software, even if advised of the possibility of
33 .Nd "performance measurement with performance monitoring hardware"
41 .Op Fl M Ar mapfilename
43 .Op Fl O Ar logfilename
44 .Op Fl P Ar event-spec
45 .Op Fl R Ar logfilename
46 .Op Fl S Ar event-spec
57 .Op Fl o Ar outputfile
58 .Op Fl p Ar event-spec
61 .Op Fl s Ar event-spec
62 .Op Fl t Ar process-spec
65 .Op Fl z Ar graphdepth
66 .Op Ar command Op Ar args
70 utility measures system performance using the facilities provided by
75 utility can measure both hardware events seen by the system as a
76 whole, and those seen when a specified set of processes are executing
78 If a specific set of processes is being targeted (for example,
81 option is specified, or if a command line is specified using
83 then measurement occurs till
85 exits, or till all target processes specified by the
87 options exit, or till the
89 utility is interrupted by the user.
90 If a specific set of processes is not targeted for measurement, then
92 will perform system-wide measurements till interrupted by the
97 can mix allocations of system-mode and process-mode PMCs, of both
98 counting and sampling flavors.
99 The values of all counting PMCs are printed in human readable form
100 at regular intervals by
102 The output of sampling PMCs may be configured to go to a log file for
103 subsequent offline analysis, or, at the expense of greater
104 overhead, may be configured to be printed in text form on the fly.
106 Hardware events to measure are specified to
108 using event specifier strings
110 The syntax of these event specifiers is machine dependent and is
114 A process-mode PMC may be configured to be inheritable by the target
115 process' current and future children.
117 The following options are available:
118 .Bl -tag -width indent
120 Toggle between showing cumulative or incremental counts for
121 subsequent counting mode PMCs specified on the command line.
122 The default is to show incremental counts.
124 Create files with per-program samples in the directory named
127 The default is to create these files in the current directory.
129 Toggle showing per-process counts at the time a tracked process
130 exits for subsequent process-mode PMCs specified on the command line.
131 This option is useful for mapping the performance characteristics of a
132 complex pipeline of processes when used in conjunction with the
135 The default is to not to enable per-process tracking.
137 Print calltree (Kcachegrind) information to file
143 this information is sent to the output file specified by the
147 Print callchain information to file
153 this information is sent to the output file specified by the
156 .It Fl M Ar mapfilename
157 Write the mapping between executable objects encountered in the event
158 log and the abbreviated pathnames used for
162 If this option is not specified, mapping information is not written.
167 in which case this mapping information is sent to the output
168 file configured by the
172 Toggle capturing callchain information for subsequent sampling PMCs.
173 The default is for sampling PMCs to capture callchain information.
174 .It Fl O Ar logfilename
175 Send logging output to file
180 .Ar hostname Ns : Ns Ar port ,
183 does not start with a
189 will open a network socket to host
196 option is not specified and one of the logging options is requested,
199 will print a textual form of the logged events to the configured
201 .It Fl P Ar event-spec
202 Allocate a process mode sampling PMC measuring hardware events
205 .It Fl R Ar logfilename
206 Perform offline analysis using sampling data in file
208 .It Fl S Ar event-spec
209 Allocate a system mode sampling PMC measuring hardware events
213 Use a top like mode for sampling PMCs. The following hotkeys
214 can be used: 'c+a' switch to accumulative mode, 'c+d' switch
215 to delta mode, 'm' merge PMCs, 'n' change view, 'p' show next
216 PMC, ' ' pause, 'q' quit. calltree only: 'f' cost under threshold
219 Toggle logging the incremental counts seen by the threads of a
220 tracked process each time they are scheduled on a CPU.
221 This is an experimental feature intended to help analyse the
222 dynamic behaviour of processes in the system.
223 It may incur substantial overhead if enabled.
224 The default is for this feature to be disabled.
226 Perform a symbol and file:line lookup for each address in each
227 callgraph and save the output to
231 that only resolves the first symbol in the graph, this resolves
232 every node in the callgraph, or prints out addresses if no
233 lookup information is available.
234 This option requires the
236 option to read in samples that were previously collected and
241 Set the cpus for subsequent system mode PMCs specified on the
246 is a comma separated list of CPU numbers, or the literal
248 denoting all unhalted CPUs.
249 The default is to allocate system mode PMCs on all unhalted
252 Toggle between process mode PMCs measuring events for the target
253 process' current and future children or only measuring events for
255 The default is to measure events for the target process alone.
256 .It Fl f Ar pluginopt
257 Pass option string to the active plugin.
259 threshold=<float> do not display cost under specified value (Top).
261 skiplink=0|1 replace node with cost under threshold by a dot (Top).
263 Produce profiles in a format compatible with
265 A separate profile file is generated for each executable object
267 Profile files are placed in sub-directories named by their PMC
269 .It Fl k Ar kerneldir
270 Set the pathname of the kernel directory to argument
272 This directory specifies where
274 should look for the kernel and its modules.
278 Print the sampled PCs with the name, the start and ending addresses
279 of the function within they live.
282 argument is mandatory and indicates where the information will be stored.
287 this information is sent to the output file specified by the
291 Set the default sampling rate for subsequent sampling mode
292 PMCs specified on the command line.
293 The default is to configure PMCs to sample the CPU's instruction
294 pointer every 65536 events.
295 .It Fl o Ar outputfile
296 Send counter readings and textual representations of logged data
299 The default is to send output to
301 when collecting live data and to
303 when processing a pre-existing logfile.
304 .It Fl p Ar event-spec
305 Allocate a process mode counting PMC measuring hardware events
311 Set the top of the filesystem hierarchy under which executables
312 are located to argument
316 .It Fl s Ar event-spec
317 Allocate a system mode counting PMC measuring hardware events
320 .It Fl t Ar process-spec
321 Attach process mode PMCs to the processes named by argument
325 may be a non-negative integer denoting a specific process id, or a
326 regular expression for selecting processes based on their command names.
330 Print the values of all counting mode PMCs or sampling mode PMCs
336 may be a fractional value.
337 The default interval is 5 seconds.
338 .It Fl z Ar graphdepth
339 When printing system-wide callgraphs, limit callgraphs to the depth
340 specified by argument
346 is specified, it is executed using
349 To perform system-wide statistical sampling on an AMD Athlon CPU with
350 samples taken every 32768 instruction retirals and data being sampled
354 .Dl "pmcstat -O sample.stat -n 32768 -S k7-retired-instructions"
358 and measure the number of data cache misses suffered
359 by it and its children every 12 seconds on an AMD Athlon, use:
360 .Dl "pmcstat -d -w 12 -p k7-dc-misses firefox"
362 To measure instructions retired for all processes named
365 .Dl "pmcstat -t '^emacs$' -p instructions"
367 To measure instructions retired for processes named
369 for a period of 10 seconds use:
370 .Dl "pmcstat -t '^emacs$' -p instructions sleep 10"
372 To count instruction tlb-misses on CPUs 0 and 2 on a Intel
373 Pentium Pro/Pentium III SMP system use:
374 .Dl "pmcstat -c 0,2 -s p6-itlb-miss"
376 To collect profiling information for a specific process with pid 1234
377 based on instruction cache misses seen by it use:
378 .Dl "pmcstat -P ic-misses -t 1234 -O /tmp/sample.out"
380 To perform system-wide sampling on all configured processors
381 based on processor instructions retired use:
382 .Dl "pmcstat -S instructions -O /tmp/sample.out"
383 If callgraph capture is not desired use:
384 .Dl "pmcstat -N -S instructions -O /tmp/sample.out"
386 To send the generated event log to a remote machine use:
387 .Dl "pmcstat -S instructions -O remotehost:port"
388 On the remote machine, the sample log can be collected using
390 .Dl "nc -l remotehost port > /tmp/sample.out"
394 compatible profiles from a sample file use:
395 .Dl "pmcstat -R /tmp/sample.out -g"
397 To print a system-wide profile with callgraphs to file
400 .Dl "pmcstat -R /tmp/sample.out -G foo.graph"
406 may issue the following diagnostic messages:
408 .It "#callchain/dubious-frames"
409 The number of callchain records that had an
411 value for a return address.
412 .It "#exec handling errors"
415 events in the log file that named executables that could not be
420 events that named ELF executables.
424 events that named executables with unrecognized formats.
426 The total number of samples in the log file.
427 .It "#samples/unclaimed"
428 The number of samples that could not be correlated to a known
429 executable object (i.e., to an executable, shared library, the
430 kernel or the runtime loader).
431 .It "#samples/unknown-object"
432 The number of samples that were associated with an executable
433 with an unrecognized object format.
438 Due to the limitations of the
442 compatible profiles generated by the
444 option do not contain information about calls that cross executable
448 files are also only meaningful for native executables.
461 utility first appeared in
466 .An Joseph Koshy Aq jkoshy@FreeBSD.org
470 utility cannot yet analyse
472 logs generated by non-native architectures.