1 .\" Copyright (c) 2003-2008 Joseph Koshy. All rights reserved.
3 .\" Redistribution and use in source and binary forms, with or without
4 .\" modification, are permitted provided that the following conditions
6 .\" 1. Redistributions of source code must retain the above copyright
7 .\" notice, this list of conditions and the following disclaimer.
8 .\" 2. Redistributions in binary form must reproduce the above copyright
9 .\" notice, this list of conditions and the following disclaimer in the
10 .\" documentation and/or other materials provided with the distribution.
12 .\" THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
13 .\" ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
14 .\" IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
15 .\" ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
16 .\" FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
17 .\" DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
18 .\" OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
19 .\" HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
20 .\" LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
21 .\" OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29 .Nd library for accessing hardware performance monitoring counters
37 provides a programming interface that allows applications to use
38 hardware performance counters to gather performance data about
39 specific processes or for the system as a whole.
40 The library is implemented using the lower-level facilities offered by
45 Performance monitoring counters (PMCs) are represented by the library
46 using a software abstraction.
49 PMCs can have two scopes:
53 These PMCs measure events in a whole-system manner, i.e., independent
54 of the currently executing thread.
55 System scope PMCs are allocated on specific CPUs and do not
57 Non-privileged process are allowed to allocate system scope PMCs if the
60 .Va security.bsd.unprivileged_syspmcs
64 These PMCs only measure hardware events when the processes they are
65 attached to are executing on a CPU.
66 In an SMP system, process scope PMCs migrate between CPUs along with
67 their target processes.
70 Orthogonal to PMC scope, PMCs may be allocated in one of two
74 Counting PMCs measure events according to their scope
76 The application needs to explicitly read these counters
77 to retrieve their value.
79 Sampling PMCs cause the CPU to be periodically interrupted
80 and information about its state of execution to be collected.
81 Sampling PMCs are used to profile specific processes and kernel
82 threads or to profile the system as a whole.
85 The scope and operational mode for a software PMC are specified at
87 An application is allowed to allocate multiple PMCs subject
88 to availability of hardware resources.
90 The library uses human-readable strings to name the event being
92 The syntax used for specifying a hardware event along with additional
93 event specific qualifiers (if any) is described in detail in section
94 .Sx "EVENT SPECIFIERS"
97 PMCs are associated with the process that allocated them and
98 will be automatically reclaimed by the system when the process exits.
99 Additionally, process-scope PMCs have to be attached to one or more
100 target processes before they can perform measurements.
101 A process-scope PMC may be attached to those target processes
102 that its owner process would otherwise be permitted to debug.
103 An owner process may attach PMCs to itself allowing
104 it to measure its own behavior.
105 Additionally, on some machine architectures, such self-attached PMCs
106 may be read cheaply using specialized instructions supported by the
109 Certain kinds of PMCs require that a log file be configured before
114 System scope sampling PMCs.
116 Process scope sampling PMCs.
118 Process scope counting PMCs that have been configured to report PMC
119 readings on process context switches or process exits.
122 Up to one log file may be configured per owner process.
123 Events logged to a log file may be subsequently analyzed using the
127 The CPUs known to the PMC library are named by the
128 .Vt "enum pmc_cputype"
130 Supported CPUs include:
132 .Bl -tag -width "Li PMC_CPU_ARMV7_CORTEX_A15" -compact
133 .It Li PMC_CPU_AMD_K7
136 .It Li PMC_CPU_AMD_K8
139 .It Li PMC_CPU_ARMV7_CORTEX_A5
143 .It Li PMC_CPU_ARMV7_CORTEX_A7
147 .It Li PMC_CPU_ARMV7_CORTEX_A8
151 .It Li PMC_CPU_ARMV7_CORTEX_A9
155 .It Li PMC_CPU_ARMV7_CORTEX_A15
158 .It Li PMC_CPU_ARMV7_CORTEX_A17
162 .It Li PMC_CPU_ARMV8_CORTEX_A53
166 .It Li PMC_CPU_ARMV8_CORTEX_A57
170 .It Li PMC_CPU_ARMV8_CORTEX_A76
176 .It Li PMC_CPU_INTEL_ATOM
179 CPUs and other CPUs conforming to version 3 of the
181 performance measurement architecture.
182 .It Li PMC_CPU_INTEL_CORE
187 CPUs, and other CPUs conforming to version 1 of the
189 performance measurement architecture.
190 .It Li PMC_CPU_INTEL_CORE2
196 CPUs, and other CPUs conforming to version 2 of the
198 performance measurement architecture.
199 .It Li PMC_CPU_PPC_7450
202 .It Li PMC_CPU_PPC_970
206 .It Li PMC_CPU_PPC_E500
209 .It Li PMC_CPU_PPC_POWER8
216 PMCs supported by this library are named by the
219 Supported PMC classes include:
221 .Bl -tag -width "Li PMC_CLASS_POWER8" -compact
223 Fixed function hardware counters presents in CPUs conforming to the
225 performance measurement architecture version 2 and later.
227 Programmable hardware counters present in CPUs conforming to the
229 performance measurement architecture version 1 and later.
231 Programmable hardware counters present in
235 Programmable hardware counters present in
239 The timestamp counter on i386 and amd64 architecture CPUs.
240 .It Li PMC_CLASS_ARMV7
242 .It Li PMC_CLASS_ARMV8
244 .It Li PMC_CLASS_PPC970
248 .It Li PMC_CLASS_POWER8
252 .It Li PMC_CLASS_SOFT
256 Capabilities of performance monitoring hardware are denoted using
260 Supported capabilities include:
262 .Bl -tag -width "Li PMC_CAP_INTERRUPT" -compact
263 .It Li PMC_CAP_CASCADE
264 The ability to cascade counters.
265 .It Li PMC_CAP_DOMWIDE
266 Separate counters tied to each NUMA domain.
268 The ability to count negated to asserted transitions of the hardware
269 conditions being probed for.
270 .It Li PMC_CAP_INTERRUPT
271 The ability to interrupt the CPU.
272 .It Li PMC_CAP_INVERT
273 The ability to invert the sense of the hardware conditions being
275 .It Li PMC_CAP_PRECISE
276 The ability to perform precise sampling.
277 .It Li PMC_CAP_QUALIFIER
278 The hardware allows monitored to be further qualified in some
279 system dependent way.
281 The ability to read from performance counters.
282 .It Li PMC_CAP_SYSTEM
283 The ability to restrict counting of hardware events to when the CPU is
284 running privileged code.
285 .It Li PMC_CAP_SYSWIDE
286 A single counter aggregating events for the whole system.
287 .It Li PMC_CAP_THRESHOLD
288 The ability to ignore simultaneous hardware events below a
289 programmable threshold.
291 The ability to restrict counting of hardware events to those when the
292 CPU is running unprivileged code.
294 The ability to write to performance counters.
296 .Ss CPU Naming Conventions
297 CPUs are named using small integers from zero up to, but
298 excluding, the value returned by function
300 On platforms supporting sparsely numbered CPUs not all the numbers in
301 this range will denote valid CPUs.
302 Operations on non-existent CPUs will return an error.
303 .Ss Functional Grouping of the API
304 This section contains a brief overview of the available functionality
306 Each function listed here is described further in its own manual page.
309 .Bl -tag -width 6n -compact
310 .It Fn pmc_disable , Fn pmc_enable
311 Administratively disable (enable) specific performance monitoring
313 Counters that are disabled will not be available to applications to
316 .It "Convenience Functions"
317 .Bl -tag -width 6n -compact
318 .It Fn pmc_event_names_of_class
319 Returns a list of event names supported by a given PMC type.
320 .It Fn pmc_name_of_capability
323 flag to a human-readable string.
324 .It Fn pmc_name_of_class
327 constant to a human-readable string.
328 .It Fn pmc_name_of_cputype
329 Return a human-readable name for a CPU type.
330 .It Fn pmc_name_of_disposition
331 Return a human-readable string describing a PMC's disposition.
332 .It Fn pmc_name_of_event
333 Convert a numeric event code to a human-readable string.
334 .It Fn pmc_name_of_mode
337 constant to a human-readable name.
338 .It Fn pmc_name_of_state
339 Return a human-readable string describing a PMC's current state.
341 .It "Library Initialization"
342 .Bl -tag -width 6n -compact
344 Initialize the library.
345 This function must be called before any other library function.
347 .It "Log File Handling"
348 .Bl -tag -width 6n -compact
349 .It Fn pmc_configure_logfile
350 Configure a log file for
352 to write logged events to.
353 .It Fn pmc_flush_logfile
354 Flush all pending log data in
357 .It Fn pmc_close_logfile
358 Flush all pending log data and close
362 Append arbitrary user data to the current log file.
365 .Bl -tag -width 6n -compact
366 .It Fn pmc_allocate , Fn pmc_release
367 Allocate (free) a PMC.
368 .It Fn pmc_attach , Fn pmc_detach
369 Attach (detach) a process scope PMC to a target.
370 .It Fn pmc_read , Fn pmc_write , Fn pmc_rw
371 Read (write) a value from (to) a PMC.
372 .It Fn pmc_start , Fn pmc_stop
373 Start (stop) a software PMC.
375 Set the reload value for a sampling PMC.
378 .Bl -tag -width 6n -compact
379 .It Fn pmc_capabilities
380 Retrieve the capabilities for a given PMC.
382 Retrieve information about the CPUs and PMC hardware present in the
384 .It Fn pmc_get_driver_stats
385 Retrieve statistics maintained by
388 Determine the greatest possible CPU number on the system.
390 Return the number of hardware PMCs present in a given CPU.
392 Return information about the state of a given CPU's PMCs.
394 Determine the width of a hardware counter in bits.
396 .It "x86 Architecture Specific API"
397 .Bl -tag -width 6n -compact
399 Returns the processor model specific register number
402 Applications may then use the x86
404 instruction to directly read the contents of the PMC.
407 .Ss Signal Handling Requirements
408 Applications using PMCs are required to handle the following signals:
409 .Bl -tag -width ".Dv SIGBUS"
413 module is unloaded using
415 processes that have PMCs allocated to them will be sent a
421 driver will send a PMC owning process a
426 any process-mode PMC allocated by it loses all its
429 the driver encounters an error when writing log data to a
431 This error may be retrieved by a subsequent call to
432 .Fn pmc_flush_logfile .
435 .Ss Typical Program Flow
438 An application would first invoke function
440 to allow the library to initialize itself.
442 Signal handling would then be set up.
444 Next the application would allocate the PMCs it desires using function
447 Initial values for PMCs may be set using function
450 If a log file is necessary for the PMCs to work, it would
451 be configured using function
452 .Fn pmc_configure_logfile .
454 Process scope PMCs would then be attached to their target processes
458 The PMCs would then be started using function
461 Once started, the values of counting PMCs may be read using function
463 For PMCs that write events to the log file, this logged data would be
464 read and parsed using the
468 PMCs are stopped using function
470 and process scope PMCs are detached from their targets using
474 Before the process exits, it may release its PMCs using function
476 Any configured log file may be closed using function
477 .Fn pmc_configure_logfile .
480 Event specifiers are strings comprising of an event name, followed by
481 optional parameters modifying the semantics of the hardware event
483 Event names are PMC architecture dependent, but the PMC library defines
484 machine independent aliases for commonly used events.
486 Event specifiers spellings are case-insensitive and space characters,
487 periods, underscores and hyphens are considered equivalent to each other.
488 Thus the event specifiers
489 .Qq "Example Event" ,
490 .Qq "example-event" ,
494 .Ss PMC Architecture Dependent Events
495 PMC architecture dependent event specifiers are described in the
496 following manual pages:
497 .Bl -column " PMC_CLASS_TSC " "MANUAL PAGE "
498 .It Em "PMC Class" Ta Em "Manual Page"
499 .It Li PMC_CLASS_IAF Ta Xr pmc.iaf 3
500 .It Li PMC_CLASS_IAP Ta Xr pmc.atom 3 , Xr pmc.core 3 , Xr pmc.core2 3
501 .It Li PMC_CLASS_K7 Ta Xr pmc.k7 3
502 .It Li PMC_CLASS_K8 Ta Xr pmc.k8 3
503 .It Li PMC_CLASS_TSC Ta Xr pmc.tsc 3
505 .Ss Event Name Aliases
506 Event name aliases are PMC-independent names for commonly used events.
507 The following aliases are known to this version of the
510 .Bl -tag -width indent
512 Measure the number of branches retired.
513 .It Li branch-mispredicts
514 Measure the number of retired branches that were mispredicted.
516 Measure processor cycles.
517 This event is implemented using the processor's Time Stamp Counter
520 Measure the number of data cache misses.
522 Measure the number of instruction cache misses.
524 Measure the number of instructions retired.
526 Measure the number of interrupts seen.
527 .It Li unhalted-cycles
528 Measure the number of cycles the processor is not in a halted
532 The interface between the
536 driver is intended to be private to the implementation and may
538 In order to ease forward compatibility with future versions of the
540 driver, applications are urged to dynamically link with the
543 Doing otherwise is unsupported.
549 .Xr pmc.haswelluc 3 ,
550 .Xr pmc.haswellxeon 3 ,
552 .Xr pmc.ivybridge 3 ,
553 .Xr pmc.ivybridgexeon 3 ,
556 .Xr pmc.sandybridge 3 ,
557 .Xr pmc.sandybridgeuc 3 ,
558 .Xr pmc.sandybridgexeon 3 ,
562 .Xr pmc.westmereuc 3 ,
565 .Xr pmc_capabilities 3 ,
566 .Xr pmc_configure_logfile 3 ,
568 .Xr pmc_event_names_of_class 3 ,
569 .Xr pmc_get_driver_stats 3 ,
572 .Xr pmc_name_of_capability 3 ,
583 library first appeared in
588 library was written by
589 .An Joseph Koshy Aq Mt jkoshy@FreeBSD.org .