tjh.dev
Linux kernel mirror (for testing)
git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
kernel
os
linux
1 i synthesize instructions events
2 b synthesize branches events (branch misses for Arm SPE)
3 c synthesize branches events (calls only)
4 r synthesize branches events (returns only)
5 x synthesize transactions events
6 w synthesize ptwrite events
7 p synthesize power events (incl. PSB events for Intel PT)
8 o synthesize other events recorded due to the use
9 of aux-output (refer to perf record)
10 e synthesize error events
11 d create a debug log
12 f synthesize first level cache events
13 m synthesize last level cache events
14 M synthesize memory events
15 t synthesize TLB events
16 a synthesize remote access events
17 g synthesize a call chain (use with i or x)
18 G synthesize a call chain on existing event records
19 l synthesize last branch entries (use with i or x)
20 L synthesize last branch entries on existing event records
21 s skip initial number of events
22 q quicker (less detailed) decoding
23 A approximate IPC
24 Z prefer to ignore timestamps (so-called "timeless" decoding)
25
26 The default is all events i.e. the same as --itrace=ibxwpe,
27 except for perf script where it is --itrace=ce
28
29 In addition, the period (default 100000, except for perf script where it is 1)
30 for instructions events can be specified in units of:
31
32 i instructions
33 t ticks
34 ms milliseconds
35 us microseconds
36 ns nanoseconds (default)
37
38 Also the call chain size (default 16, max. 1024) for instructions or
39 transactions events can be specified.
40
41 Also the number of last branch entries (default 64, max. 1024) for
42 instructions or transactions events can be specified.
43
44 Similar to options g and l, size may also be specified for options G and L.
45 On x86, note that G and L work poorly when data has been recorded with
46 large PEBS. Refer linkperf:perf-intel-pt[1] man page for details.
47
48 It is also possible to skip events generated (instructions, branches, transactions,
49 ptwrite, power) at the beginning. This is useful to ignore initialization code.
50
51 --itrace=i0nss1000000
52
53 skips the first million instructions.
54
55 The 'e' option may be followed by flags which affect what errors will or
56 will not be reported. Each flag must be preceded by either '+' or '-'.
57 The flags are:
58 o overflow
59 l trace data lost
60
61 If supported, the 'd' option may be followed by flags which affect what
62 debug messages will or will not be logged. Each flag must be preceded
63 by either '+' or '-'. The flags are:
64 a all perf events
65 o output to stdout
66
67 If supported, the 'q' option may be repeated to increase the effect.