xref: /OK3568_Linux_fs/kernel/tools/perf/Documentation/itrace.txt (revision 4882a59341e53eb6f0b4789bf948001014eff981)
1*4882a593Smuzhiyun		i	synthesize instructions events
2*4882a593Smuzhiyun		b	synthesize branches events (branch misses for Arm SPE)
3*4882a593Smuzhiyun		c	synthesize branches events (calls only)
4*4882a593Smuzhiyun		r	synthesize branches events (returns only)
5*4882a593Smuzhiyun		x	synthesize transactions events
6*4882a593Smuzhiyun		w	synthesize ptwrite events
7*4882a593Smuzhiyun		p	synthesize power events
8*4882a593Smuzhiyun		o	synthesize other events recorded due to the use
9*4882a593Smuzhiyun			of aux-output (refer to perf record)
10*4882a593Smuzhiyun		e	synthesize error events
11*4882a593Smuzhiyun		d	create a debug log
12*4882a593Smuzhiyun		f	synthesize first level cache events
13*4882a593Smuzhiyun		m	synthesize last level cache events
14*4882a593Smuzhiyun		t	synthesize TLB events
15*4882a593Smuzhiyun		a	synthesize remote access events
16*4882a593Smuzhiyun		g	synthesize a call chain (use with i or x)
17*4882a593Smuzhiyun		G	synthesize a call chain on existing event records
18*4882a593Smuzhiyun		l	synthesize last branch entries (use with i or x)
19*4882a593Smuzhiyun		L	synthesize last branch entries on existing event records
20*4882a593Smuzhiyun		s       skip initial number of events
21*4882a593Smuzhiyun		q	quicker (less detailed) decoding
22*4882a593Smuzhiyun
23*4882a593Smuzhiyun	The default is all events i.e. the same as --itrace=ibxwpe,
24*4882a593Smuzhiyun	except for perf script where it is --itrace=ce
25*4882a593Smuzhiyun
26*4882a593Smuzhiyun	In addition, the period (default 100000, except for perf script where it is 1)
27*4882a593Smuzhiyun	for instructions events can be specified in units of:
28*4882a593Smuzhiyun
29*4882a593Smuzhiyun		i	instructions
30*4882a593Smuzhiyun		t	ticks
31*4882a593Smuzhiyun		ms	milliseconds
32*4882a593Smuzhiyun		us	microseconds
33*4882a593Smuzhiyun		ns	nanoseconds (default)
34*4882a593Smuzhiyun
35*4882a593Smuzhiyun	Also the call chain size (default 16, max. 1024) for instructions or
36*4882a593Smuzhiyun	transactions events can be specified.
37*4882a593Smuzhiyun
38*4882a593Smuzhiyun	Also the number of last branch entries (default 64, max. 1024) for
39*4882a593Smuzhiyun	instructions or transactions events can be specified.
40*4882a593Smuzhiyun
41*4882a593Smuzhiyun	Similar to options g and l, size may also be specified for options G and L.
42*4882a593Smuzhiyun	On x86, note that G and L work poorly when data has been recorded with
43*4882a593Smuzhiyun	large PEBS. Refer linkperf:perf-intel-pt[1] man page for details.
44*4882a593Smuzhiyun
45*4882a593Smuzhiyun	It is also possible to skip events generated (instructions, branches, transactions,
46*4882a593Smuzhiyun	ptwrite, power) at the beginning. This is useful to ignore initialization code.
47*4882a593Smuzhiyun
48*4882a593Smuzhiyun	--itrace=i0nss1000000
49*4882a593Smuzhiyun
50*4882a593Smuzhiyun	skips the first million instructions.
51*4882a593Smuzhiyun
52*4882a593Smuzhiyun	The 'e' option may be followed by flags which affect what errors will or
53*4882a593Smuzhiyun	will not be reported. Each flag must be preceded by either '+' or '-'.
54*4882a593Smuzhiyun	The flags are:
55*4882a593Smuzhiyun		o	overflow
56*4882a593Smuzhiyun		l	trace data lost
57*4882a593Smuzhiyun
58*4882a593Smuzhiyun	If supported, the 'd' option may be followed by flags which affect what
59*4882a593Smuzhiyun	debug messages will or will not be logged. Each flag must be preceded
60*4882a593Smuzhiyun	by either '+' or '-'. The flags are:
61*4882a593Smuzhiyun		a	all perf events
62*4882a593Smuzhiyun
63*4882a593Smuzhiyun	If supported, the 'q' option may be repeated to increase the effect.
64