1perf-list(1)
2============
3
4NAME
5----
6perf-list - List all symbolic event types
7
8SYNOPSIS
9--------
10[verse]
11'perf list' [--no-desc] [--long-desc]
12            [hw|sw|cache|tracepoint|pmu|sdt|metric|metricgroup|event_glob]
13
14DESCRIPTION
15-----------
16This command displays the symbolic event types which can be selected in the
17various perf commands with the -e option.
18
19OPTIONS
20-------
21-d::
22--desc::
23Print extra event descriptions. (default)
24
25--no-desc::
26Don't print descriptions.
27
28-v::
29--long-desc::
30Print longer event descriptions.
31
32--debug::
33Enable debugging output.
34
35--details::
36Print how named events are resolved internally into perf events, and also
37any extra expressions computed by perf stat.
38
39--deprecated::
40Print deprecated events. By default the deprecated events are hidden.
41
42[[EVENT_MODIFIERS]]
43EVENT MODIFIERS
44---------------
45
46Events can optionally have a modifier by appending a colon and one or
47more modifiers. Modifiers allow the user to restrict the events to be
48counted. The following modifiers exist:
49
50 u - user-space counting
51 k - kernel counting
52 h - hypervisor counting
53 I - non idle counting
54 G - guest counting (in KVM guests)
55 H - host counting (not in KVM guests)
56 p - precise level
57 P - use maximum detected precise level
58 S - read sample value (PERF_SAMPLE_READ)
59 D - pin the event to the PMU
60 W - group is weak and will fallback to non-group if not schedulable,
61
62The 'p' modifier can be used for specifying how precise the instruction
63address should be. The 'p' modifier can be specified multiple times:
64
65 0 - SAMPLE_IP can have arbitrary skid
66 1 - SAMPLE_IP must have constant skid
67 2 - SAMPLE_IP requested to have 0 skid
68 3 - SAMPLE_IP must have 0 skid, or uses randomization to avoid
69     sample shadowing effects.
70
71For Intel systems precise event sampling is implemented with PEBS
72which supports up to precise-level 2, and precise level 3 for
73some special cases
74
75On AMD systems it is implemented using IBS (up to precise-level 2).
76The precise modifier works with event types 0x76 (cpu-cycles, CPU
77clocks not halted) and 0xC1 (micro-ops retired). Both events map to
78IBS execution sampling (IBS op) with the IBS Op Counter Control bit
79(IbsOpCntCtl) set respectively (see AMD64 Architecture Programmer’s
80Manual Volume 2: System Programming, 13.3 Instruction-Based
81Sampling). Examples to use IBS:
82
83 perf record -a -e cpu-cycles:p ...    # use ibs op counting cycles
84 perf record -a -e r076:p ...          # same as -e cpu-cycles:p
85 perf record -a -e r0C1:p ...          # use ibs op counting micro-ops
86
87RAW HARDWARE EVENT DESCRIPTOR
88-----------------------------
89Even when an event is not available in a symbolic form within perf right now,
90it can be encoded in a per processor specific way.
91
92For instance For x86 CPUs NNN represents the raw register encoding with the
93layout of IA32_PERFEVTSELx MSRs (see [Intel® 64 and IA-32 Architectures Software Developer's Manual Volume 3B: System Programming Guide] Figure 30-1 Layout
94of IA32_PERFEVTSELx MSRs) or AMD's PerfEvtSeln (see [AMD64 Architecture Programmer’s Manual Volume 2: System Programming], Page 344,
95Figure 13-7 Performance Event-Select Register (PerfEvtSeln)).
96
97Note: Only the following bit fields can be set in x86 counter
98registers: event, umask, edge, inv, cmask. Esp. guest/host only and
99OS/user mode flags must be setup using <<EVENT_MODIFIERS, EVENT
100MODIFIERS>>.
101
102Example:
103
104If the Intel docs for a QM720 Core i7 describe an event as:
105
106  Event  Umask  Event Mask
107  Num.   Value  Mnemonic    Description                        Comment
108
109  A8H      01H  LSD.UOPS    Counts the number of micro-ops     Use cmask=1 and
110                            delivered by loop stream detector  invert to count
111                                                               cycles
112
113raw encoding of 0x1A8 can be used:
114
115 perf stat -e r1a8 -a sleep 1
116 perf record -e r1a8 ...
117
118It's also possible to use pmu syntax:
119
120 perf record -e r1a8 -a sleep 1
121 perf record -e cpu/r1a8/ ...
122
123You should refer to the processor specific documentation for getting these
124details. Some of them are referenced in the SEE ALSO section below.
125
126ARBITRARY PMUS
127--------------
128
129perf also supports an extended syntax for specifying raw parameters
130to PMUs. Using this typically requires looking up the specific event
131in the CPU vendor specific documentation.
132
133The available PMUs and their raw parameters can be listed with
134
135  ls /sys/devices/*/format
136
137For example the raw event "LSD.UOPS" core pmu event above could
138be specified as
139
140  perf stat -e cpu/event=0xa8,umask=0x1,name=LSD.UOPS_CYCLES,cmask=0x1/ ...
141
142  or using extended name syntax
143
144  perf stat -e cpu/event=0xa8,umask=0x1,cmask=0x1,name=\'LSD.UOPS_CYCLES:cmask=0x1\'/ ...
145
146PER SOCKET PMUS
147---------------
148
149Some PMUs are not associated with a core, but with a whole CPU socket.
150Events on these PMUs generally cannot be sampled, but only counted globally
151with perf stat -a. They can be bound to one logical CPU, but will measure
152all the CPUs in the same socket.
153
154This example measures memory bandwidth every second
155on the first memory controller on socket 0 of a Intel Xeon system
156
157  perf stat -C 0 -a uncore_imc_0/cas_count_read/,uncore_imc_0/cas_count_write/ -I 1000 ...
158
159Each memory controller has its own PMU.  Measuring the complete system
160bandwidth would require specifying all imc PMUs (see perf list output),
161and adding the values together. To simplify creation of multiple events,
162prefix and glob matching is supported in the PMU name, and the prefix
163'uncore_' is also ignored when performing the match. So the command above
164can be expanded to all memory controllers by using the syntaxes:
165
166  perf stat -C 0 -a imc/cas_count_read/,imc/cas_count_write/ -I 1000 ...
167  perf stat -C 0 -a *imc*/cas_count_read/,*imc*/cas_count_write/ -I 1000 ...
168
169This example measures the combined core power every second
170
171  perf stat -I 1000 -e power/energy-cores/  -a
172
173ACCESS RESTRICTIONS
174-------------------
175
176For non root users generally only context switched PMU events are available.
177This is normally only the events in the cpu PMU, the predefined events
178like cycles and instructions and some software events.
179
180Other PMUs and global measurements are normally root only.
181Some event qualifiers, such as "any", are also root only.
182
183This can be overridden by setting the kernel.perf_event_paranoid
184sysctl to -1, which allows non root to use these events.
185
186For accessing trace point events perf needs to have read access to
187/sys/kernel/debug/tracing, even when perf_event_paranoid is in a relaxed
188setting.
189
190TRACING
191-------
192
193Some PMUs control advanced hardware tracing capabilities, such as Intel PT,
194that allows low overhead execution tracing.  These are described in a separate
195intel-pt.txt document.
196
197PARAMETERIZED EVENTS
198--------------------
199
200Some pmu events listed by 'perf-list' will be displayed with '?' in them. For
201example:
202
203  hv_gpci/dtbp_ptitc,phys_processor_idx=?/
204
205This means that when provided as an event, a value for '?' must
206also be supplied. For example:
207
208  perf stat -C 0 -e 'hv_gpci/dtbp_ptitc,phys_processor_idx=0x2/' ...
209
210EVENT QUALIFIERS:
211
212It is also possible to add extra qualifiers to an event:
213
214percore:
215
216Sums up the event counts for all hardware threads in a core, e.g.:
217
218
219  perf stat -e cpu/event=0,umask=0x3,percore=1/
220
221
222EVENT GROUPS
223------------
224
225Perf supports time based multiplexing of events, when the number of events
226active exceeds the number of hardware performance counters. Multiplexing
227can cause measurement errors when the workload changes its execution
228profile.
229
230When metrics are computed using formulas from event counts, it is useful to
231ensure some events are always measured together as a group to minimize multiplexing
232errors. Event groups can be specified using { }.
233
234  perf stat -e '{instructions,cycles}' ...
235
236The number of available performance counters depend on the CPU. A group
237cannot contain more events than available counters.
238For example Intel Core CPUs typically have four generic performance counters
239for the core, plus three fixed counters for instructions, cycles and
240ref-cycles. Some special events have restrictions on which counter they
241can schedule, and may not support multiple instances in a single group.
242When too many events are specified in the group some of them will not
243be measured.
244
245Globally pinned events can limit the number of counters available for
246other groups. On x86 systems, the NMI watchdog pins a counter by default.
247The nmi watchdog can be disabled as root with
248
249	echo 0 > /proc/sys/kernel/nmi_watchdog
250
251Events from multiple different PMUs cannot be mixed in a group, with
252some exceptions for software events.
253
254LEADER SAMPLING
255---------------
256
257perf also supports group leader sampling using the :S specifier.
258
259  perf record -e '{cycles,instructions}:S' ...
260  perf report --group
261
262Normally all events in an event group sample, but with :S only
263the first event (the leader) samples, and it only reads the values of the
264other events in the group.
265
266However, in the case AUX area events (e.g. Intel PT or CoreSight), the AUX
267area event must be the leader, so then the second event samples, not the first.
268
269OPTIONS
270-------
271
272Without options all known events will be listed.
273
274To limit the list use:
275
276. 'hw' or 'hardware' to list hardware events such as cache-misses, etc.
277
278. 'sw' or 'software' to list software events such as context switches, etc.
279
280. 'cache' or 'hwcache' to list hardware cache events such as L1-dcache-loads, etc.
281
282. 'tracepoint' to list all tracepoint events, alternatively use
283  'subsys_glob:event_glob' to filter by tracepoint subsystems such as sched,
284  block, etc.
285
286. 'pmu' to print the kernel supplied PMU events.
287
288. 'sdt' to list all Statically Defined Tracepoint events.
289
290. 'metric' to list metrics
291
292. 'metricgroup' to list metricgroups with metrics.
293
294. If none of the above is matched, it will apply the supplied glob to all
295  events, printing the ones that match.
296
297. As a last resort, it will do a substring search in all event names.
298
299One or more types can be used at the same time, listing the events for the
300types specified.
301
302Support raw format:
303
304. '--raw-dump', shows the raw-dump of all the events.
305. '--raw-dump [hw|sw|cache|tracepoint|pmu|event_glob]', shows the raw-dump of
306  a certain kind of events.
307
308SEE ALSO
309--------
310linkperf:perf-stat[1], linkperf:perf-top[1],
311linkperf:perf-record[1],
312http://www.intel.com/sdm/[Intel® 64 and IA-32 Architectures Software Developer's Manual Volume 3B: System Programming Guide],
313http://support.amd.com/us/Processor_TechDocs/24593_APM_v2.pdf[AMD64 Architecture Programmer’s Manual Volume 2: System Programming]
314