1PERF-INJECT(1) perf Manual PERF-INJECT(1)
2
6 perf-inject - Filter to augment the events stream with additional
7 information
8
10 perf inject <options>
11
13 perf-inject reads a perf-record event stream and repipes it to stdout.
14 At any point the processing code can inject other events into the event
15 stream - in this case build-ids (-b option) are read and injected as
16 needed into the event stream.
17 Build-ids are just the first user of perf-inject - potentially anything
19 that needs userspace processing to augment the events stream with
20 additional information could make use of this facility.
21
23 -b, --build-ids
24 Inject build-ids into the output stream
25 --buildid-all: Inject build-ids of all DSOs into the output stream
27 -v, --verbose
29 Be more verbose.
30 -i, --input=
32 Input file name. (default: stdin)
33 -o, --output=
35 Output file name. (default: stdout)
36 -s, --sched-stat
38 Merge sched_stat and sched_switch for getting events where and how
39 long tasks slept. sched_switch contains a callchain where a task
40 slept and sched_stat contains a timeslice how long a task slept.
41 -k, --vmlinux=<file>
43 vmlinux pathname
44 --ignore-vmlinux
46 Ignore vmlinux files.
47 --kallsyms=<file>
49 kallsyms pathname
50 --itrace
52 Decode Instruction Tracing data, replacing it with synthesized
53 events. Options are:
54 i synthesize instructions events
56 b synthesize branches events (branch misses for Arm SPE)
57 c synthesize branches events (calls only)
58 r synthesize branches events (returns only)
59 x synthesize transactions events
60 w synthesize ptwrite events
61 p synthesize power events (incl. PSB events for Intel PT)
62 o synthesize other events recorded due to the use
63 of aux-output (refer to perf record)
64 I synthesize interrupt or similar (asynchronous) events
65 (e.g. Intel PT Event Trace)
66 e synthesize error events
67 d create a debug log
68 f synthesize first level cache events
69 m synthesize last level cache events
70 M synthesize memory events
71 t synthesize TLB events
72 a synthesize remote access events
73 g synthesize a call chain (use with i or x)
74 G synthesize a call chain on existing event records
75 l synthesize last branch entries (use with i or x)
76 L synthesize last branch entries on existing event records
77 s skip initial number of events
78 q quicker (less detailed) decoding
79 A approximate IPC
80 Z prefer to ignore timestamps (so-called "timeless" decoding)
81 The default is all events i.e. the same as --itrace=ibxwpe,
83 except for perf script where it is --itrace=ce
84 In addition, the period (default 100000, except for perf script where it is 1)
86 for instructions events can be specified in units of:
87 i instructions
89 t ticks
90 ms milliseconds
91 us microseconds
92 ns nanoseconds (default)
93 Also the call chain size (default 16, max. 1024) for instructions or
95 transactions events can be specified.
96 Also the number of last branch entries (default 64, max. 1024) for
98 instructions or transactions events can be specified.
99 Similar to options g and l, size may also be specified for options G and L.
101 On x86, note that G and L work poorly when data has been recorded with
102 large PEBS. Refer linkperf:perf-intel-pt[1] man page for details.
103 It is also possible to skip events generated (instructions, branches, transactions,
105 ptwrite, power) at the beginning. This is useful to ignore initialization code.
106 --itrace=i0nss1000000
108 skips the first million instructions.
110 The 'e' option may be followed by flags which affect what errors will or
112 will not be reported. Each flag must be preceded by either '+' or '-'.
113 The flags are:
114 o overflow
115 l trace data lost
116 If supported, the 'd' option may be followed by flags which affect what
118 debug messages will or will not be logged. Each flag must be preceded
119 by either '+' or '-'. The flags are:
120 a all perf events
121 o output to stdout
122 If supported, the 'q' option may be repeated to increase the effect.
124 --strip
126 Use with --itrace to strip out non-synthesized events.
127 -j, --jit
129 Process jitdump files by injecting the mmap records corresponding
130 to jitted functions. This option also generates the ELF images for
131 each jitted function found in the jitdumps files captured in the
132 input perf.data file. Use this option if you are monitoring
133 environment using JIT runtimes, such as Java, DART or V8.
134 -f, --force
136 Don’t complain, do it.
137 --vm-time-correlation[=OPTIONS]
139 Some architectures may capture AUX area data which contains
140 timestamps affected by virtualization. This option will update
141 those timestamps in place, to correlate with host timestamps. The
142 in-place update means that an output file is not specified, and
143 instead the input file is modified. The options are architecture
144 specific, except that they may start with "dry-run" which will
145 cause the file to be processed but without updating it. Currently
146 this option is supported only by Intel PT, refer perf-intel-pt(1)
147
149 perf-record(1), perf-report(1), perf-archive(1), perf-intel-pt(1)
150perf 06/14/2022 PERF-INJECT(1)