6 perf-list - List all symbolic event types
11 'perf list' [<options>]
12 [hw|sw|cache|tracepoint|pmu|sdt|metric|metricgroup|event_glob]
16 This command displays the symbolic event types which can be selected in the
17 various perf commands with the -e option.
23 Print extra event descriptions. (default)
26 Don't print descriptions.
30 Print longer event descriptions.
33 Enable debugging output.
36 Print how named events are resolved internally into perf events, and also
37 any extra expressions computed by perf stat.
40 Print deprecated events. By default the deprecated events are hidden.
43 Print PMU events and metrics limited to the specific PMU name.
44 (e.g. --unit cpu, --unit msr, --unit cpu_core, --unit cpu_atom)
48 Output in JSON format.
52 Output file name. By default output is written to stdout.
58 Events can optionally have a modifier by appending a colon and one or
59 more modifiers. Modifiers allow the user to restrict the events to be
60 counted. The following modifiers exist:
62 u - user-space counting
64 h - hypervisor counting
66 G - guest counting (in KVM guests)
67 H - host counting (not in KVM guests)
69 P - use maximum detected precise level
70 S - read sample value (PERF_SAMPLE_READ)
71 D - pin the event to the PMU
72 W - group is weak and will fallback to non-group if not schedulable,
73 e - group or event are exclusive and do not share the PMU
74 b - use BPF aggregration (see perf stat --bpf-counters)
75 R - retire latency value of the event
77 The 'p' modifier can be used for specifying how precise the instruction
78 address should be. The 'p' modifier can be specified multiple times:
80 0 - SAMPLE_IP can have arbitrary skid
81 1 - SAMPLE_IP must have constant skid
82 2 - SAMPLE_IP requested to have 0 skid
83 3 - SAMPLE_IP must have 0 skid, or uses randomization to avoid
84 sample shadowing effects.
86 For Intel systems precise event sampling is implemented with PEBS
87 which supports up to precise-level 2, and precise level 3 for
90 On AMD systems it is implemented using IBS OP (up to precise-level 2).
91 Unlike Intel PEBS which provides levels of precision, AMD core pmu is
92 inherently non-precise and IBS is inherently precise. (i.e. ibs_op//,
93 ibs_op//p, ibs_op//pp and ibs_op//ppp are all same). The precise modifier
94 works with event types 0x76 (cpu-cycles, CPU clocks not halted) and 0xC1
95 (micro-ops retired). Both events map to IBS execution sampling (IBS op)
96 with the IBS Op Counter Control bit (IbsOpCntCtl) set respectively (see the
97 Core Complex (CCX) -> Processor x86 Core -> Instruction Based Sampling (IBS)
98 section of the [AMD Processor Programming Reference (PPR)] relevant to the
99 family, model and stepping of the processor being used).
101 Manual Volume 2: System Programming, 13.3 Instruction-Based
102 Sampling). Examples to use IBS:
104 perf record -a -e cpu-cycles:p ... # use ibs op counting cycles
105 perf record -a -e r076:p ... # same as -e cpu-cycles:p
106 perf record -a -e r0C1:p ... # use ibs op counting micro-ops
108 RAW HARDWARE EVENT DESCRIPTOR
109 -----------------------------
110 Even when an event is not available in a symbolic form within perf right now,
111 it can be encoded in a per processor specific way.
113 For instance on x86 CPUs, N is a hexadecimal value that represents the raw register encoding with the
114 layout of IA32_PERFEVTSELx MSRs (see [IntelĀ® 64 and IA-32 Architectures Software Developer's Manual Volume 3B: System Programming Guide] Figure 30-1 Layout
115 of IA32_PERFEVTSELx MSRs) or AMD's PERF_CTL MSRs (see the
116 Core Complex (CCX) -> Processor x86 Core -> MSR Registers section of the
117 [AMD Processor Programming Reference (PPR)] relevant to the family, model
118 and stepping of the processor being used).
120 Note: Only the following bit fields can be set in x86 counter
121 registers: event, umask, edge, inv, cmask. Esp. guest/host only and
122 OS/user mode flags must be setup using <<EVENT_MODIFIERS, EVENT
127 If the Intel docs for a QM720 Core i7 describe an event as:
129 Event Umask Event Mask
130 Num. Value Mnemonic Description Comment
132 A8H 01H LSD.UOPS Counts the number of micro-ops Use cmask=1 and
133 delivered by loop stream detector invert to count
136 raw encoding of 0x1A8 can be used:
138 perf stat -e r1a8 -a sleep 1
139 perf record -e r1a8 ...
141 It's also possible to use pmu syntax:
143 perf record -e r1a8 -a sleep 1
144 perf record -e cpu/r1a8/ ...
145 perf record -e cpu/r0x1a8/ ...
147 Some processors, like those from AMD, support event codes and unit masks
148 larger than a byte. In such cases, the bits corresponding to the event
149 configuration parameters can be seen with:
151 cat /sys/bus/event_source/devices/<pmu>/format/<config>
155 If the AMD docs for an EPYC 7713 processor describe an event as:
157 Event Umask Event Mask
158 Num. Value Mnemonic Description
160 28FH 03H op_cache_hit_miss.op_cache_hit Counts Op Cache micro-tag
163 raw encoding of 0x0328F cannot be used since the upper nibble of the
164 EventSelect bits have to be specified via bits 32-35 as can be seen with:
166 cat /sys/bus/event_source/devices/cpu/format/event
168 raw encoding of 0x20000038F should be used instead:
170 perf stat -e r20000038f -a sleep 1
171 perf record -e r20000038f ...
173 It's also possible to use pmu syntax:
175 perf record -e r20000038f -a sleep 1
176 perf record -e cpu/r20000038f/ ...
177 perf record -e cpu/r0x20000038f/ ...
179 You should refer to the processor specific documentation for getting these
180 details. Some of them are referenced in the SEE ALSO section below.
185 perf also supports an extended syntax for specifying raw parameters
186 to PMUs. Using this typically requires looking up the specific event
187 in the CPU vendor specific documentation.
189 The available PMUs and their raw parameters can be listed with
191 ls /sys/devices/*/format
193 For example the raw event "LSD.UOPS" core pmu event above could
196 perf stat -e cpu/event=0xa8,umask=0x1,name=LSD.UOPS_CYCLES,cmask=0x1/ ...
198 or using extended name syntax
200 perf stat -e cpu/event=0xa8,umask=0x1,cmask=0x1,name=\'LSD.UOPS_CYCLES:cmask=0x1\'/ ...
205 Some PMUs are not associated with a core, but with a whole CPU socket.
206 Events on these PMUs generally cannot be sampled, but only counted globally
207 with perf stat -a. They can be bound to one logical CPU, but will measure
208 all the CPUs in the same socket.
210 This example measures memory bandwidth every second
211 on the first memory controller on socket 0 of a Intel Xeon system
213 perf stat -C 0 -a uncore_imc_0/cas_count_read/,uncore_imc_0/cas_count_write/ -I 1000 ...
215 Each memory controller has its own PMU. Measuring the complete system
216 bandwidth would require specifying all imc PMUs (see perf list output),
217 and adding the values together. To simplify creation of multiple events,
218 prefix and glob matching is supported in the PMU name, and the prefix
219 'uncore_' is also ignored when performing the match. So the command above
220 can be expanded to all memory controllers by using the syntaxes:
222 perf stat -C 0 -a imc/cas_count_read/,imc/cas_count_write/ -I 1000 ...
223 perf stat -C 0 -a *imc*/cas_count_read/,*imc*/cas_count_write/ -I 1000 ...
225 This example measures the combined core power every second
227 perf stat -I 1000 -e power/energy-cores/ -a
232 For non root users generally only context switched PMU events are available.
233 This is normally only the events in the cpu PMU, the predefined events
234 like cycles and instructions and some software events.
236 Other PMUs and global measurements are normally root only.
237 Some event qualifiers, such as "any", are also root only.
239 This can be overridden by setting the kernel.perf_event_paranoid
240 sysctl to -1, which allows non root to use these events.
242 For accessing trace point events perf needs to have read access to
243 /sys/kernel/tracing, even when perf_event_paranoid is in a relaxed
249 Some events don't have an associated PMU instead reading values
250 available to software without perf_event_open. As these events don't
251 support sampling they can only really be read by tools like perf stat.
253 Tool events provide times and certain system parameters. Examples
254 include duration_time, user_time, system_time and num_cpus_online.
256 Hwmon events provide easy access to hwmon sysfs data typically in
257 /sys/class/hwmon. This information includes temperatures, fan speeds
264 Some PMUs control advanced hardware tracing capabilities, such as Intel PT,
265 that allows low overhead execution tracing. These are described in a separate
266 intel-pt.txt document.
271 Some pmu events listed by 'perf-list' will be displayed with '?' in them. For
274 hv_gpci/dtbp_ptitc,phys_processor_idx=?/
276 This means that when provided as an event, a value for '?' must
277 also be supplied. For example:
279 perf stat -C 0 -e 'hv_gpci/dtbp_ptitc,phys_processor_idx=0x2/' ...
283 It is also possible to add extra qualifiers to an event:
287 Sums up the event counts for all hardware threads in a core, e.g.:
290 perf stat -e cpu/event=0,umask=0x3,percore=1/
296 Perf supports time based multiplexing of events, when the number of events
297 active exceeds the number of hardware performance counters. Multiplexing
298 can cause measurement errors when the workload changes its execution
301 When metrics are computed using formulas from event counts, it is useful to
302 ensure some events are always measured together as a group to minimize multiplexing
303 errors. Event groups can be specified using { }.
305 perf stat -e '{instructions,cycles}' ...
307 The number of available performance counters depend on the CPU. A group
308 cannot contain more events than available counters.
309 For example Intel Core CPUs typically have four generic performance counters
310 for the core, plus three fixed counters for instructions, cycles and
311 ref-cycles. Some special events have restrictions on which counter they
312 can schedule, and may not support multiple instances in a single group.
313 When too many events are specified in the group some of them will not
316 Globally pinned events can limit the number of counters available for
317 other groups. On x86 systems, the NMI watchdog pins a counter by default.
318 The nmi watchdog can be disabled as root with
320 echo 0 > /proc/sys/kernel/nmi_watchdog
322 Events from multiple different PMUs cannot be mixed in a group, with
323 some exceptions for software events.
328 perf also supports group leader sampling using the :S specifier.
330 perf record -e '{cycles,instructions}:S' ...
333 Normally all events in an event group sample, but with :S only
334 the first event (the leader) samples, and it only reads the values of the
335 other events in the group.
337 However, in the case AUX area events (e.g. Intel PT or CoreSight), the AUX
338 area event must be the leader, so then the second event samples, not the first.
343 Without options all known events will be listed.
345 To limit the list use:
347 . 'hw' or 'hardware' to list hardware events such as cache-misses, etc.
349 . 'sw' or 'software' to list software events such as context switches, etc.
351 . 'cache' or 'hwcache' to list hardware cache events such as L1-dcache-loads, etc.
353 . 'tracepoint' to list all tracepoint events, alternatively use
354 'subsys_glob:event_glob' to filter by tracepoint subsystems such as sched,
357 . 'pmu' to print the kernel supplied PMU events.
359 . 'sdt' to list all Statically Defined Tracepoint events.
361 . 'metric' to list metrics
363 . 'metricgroup' to list metricgroups with metrics.
365 . If none of the above is matched, it will apply the supplied glob to all
366 events, printing the ones that match.
368 . As a last resort, it will do a substring search in all event names.
370 One or more types can be used at the same time, listing the events for the
375 . '--raw-dump', shows the raw-dump of all the events.
376 . '--raw-dump [hw|sw|cache|tracepoint|pmu|event_glob]', shows the raw-dump of
377 a certain kind of events.
381 linkperf:perf-stat[1], linkperf:perf-top[1],
382 linkperf:perf-record[1],
383 http://www.intel.com/sdm/[IntelĀ® 64 and IA-32 Architectures Software Developer's Manual Volume 3B: System Programming Guide],
384 https://bugzilla.kernel.org/show_bug.cgi?id=206537[AMD Processor Programming Reference (PPR)]