1 .TH CPUPOWER\-MONITOR "1" "22/02/2011" "" "cpu
2 .SH NAME
3 cpupower\-monitor \- Report processor frequenc
4 .SH SYNOPSIS
5 .ft B
6 .B cpupower monitor
7 .RB "\-l"
8
9 .B cpupower monitor
10 .RB [ -c ] [ "\-m <mon1>," [ "<mon2>,..." ] ]
11 .RB [ "\-i seconds" ]
12 .br
13 .B cpupower monitor
14 .RB [ -c ][ "\-m <mon1>," [ "<mon2>,..." ] ]
15 .RB command
16 .br
17 .SH DESCRIPTION
18 \fBcpupower-monitor \fP reports processor topo
19 state statistics. Either \fBcommand\fP is fork
20 statistics are printed upon its completion, or
21
22 \fBcpupower-monitor \fP implements independent
23 frequency counters. Some are retrieved from ke
24 directly reading out hardware registers. Use \
25 supported on your system.
26
27 .SH Options
28 .PP
29 \-l
30 .RS 4
31 List available monitors on your system. Additi
32 are shown:
33 .RS 2
34 .IP \(bu
35 The name in quotation marks which can be passe
36 .IP \(bu
37 The number of different counters the monitor s
38 .IP \(bu
39 The amount of time in seconds the counters mig
40 implementation constraints.
41 .IP \(bu
42 The name and a description of each counter and
43 coverage in square brackets:
44 .RS 4
45 .IP \(bu
46 [T] \-> Thread
47 .IP \(bu
48 [C] \-> Core
49 .IP \(bu
50 [P] \-> Processor Package (Socket)
51 .IP \(bu
52 [M] \-> Machine/Platform wide counter
53 .RE
54 .RE
55 .RE
56 .PP
57 \-m <mon1>,<mon2>,...
58 .RS 4
59 Only display specific monitors. Use the monito
60 .RE
61 .PP
62 \-i seconds
63 .RS 4
64 Measure interval.
65 .RE
66 .PP
67 \-c
68 .RS 4
69 Schedule the process on every core before star
70 This could be needed for the Idle_Stats monito
71 monitor (has to be run on the core that is mea
72 This is to wake up the processors from deeper
73 kernel re
74 -account its cpuidle (C-state) information bef
75 cpuidle timings from sysfs.
76 .RE
77 .PP
78 command
79 .RS 4
80 Measure idle and frequency characteristics of
81 The executable \fBcommand\fP is forked and upo
82 forked are displayed.
83 .RE
84
85 .SH MONITOR DESCRIPTIONS
86 .SS "Idle_Stats"
87 Shows statistics of the cpuidle kernel subsyst
88 /sys/devices/system/cpu/cpu*/cpuidle/state*/.
89 The kernel updates these values every time an
90 left. Therefore there can be some inaccuracy w
91 state for some time when the measure starts or
92 that one core stayed in an idle state for the
93 state usage time as exported by the kernel did
94 a state residency of 0 percent is shown while
95
96 .SS "Mperf"
97 The name comes from the aperf/mperf (average a
98 which are available on recent X86 processors.
99 (including boost frequencies).
100 The fact that on all recent hardware the mperf
101 state it is also used to show C0 (processor is
102 any sleep state) times. These counters do not
103 the "Idle_Stats" counters may show.
104 May work poorly on Linux-2.6.20 through 2.6.29
105 kernel frequency driver periodically cleared a
106 kernels.
107
108 .SS "Nehalem" "SandyBridge" "HaswellExtended"
109 Intel Core and Package sleep state counters.
110 Threads (hyperthreaded cores) may not be able
111 its sibling is utilized.
112 Deepest package sleep states may in reality sh
113 sleep states and can only be entered if all co
114 manuals (some are provided in the References s
115 The monitors are named after the CPU family wh
116 got introduced and may not match exactly the C
117 For example an IvyBridge processor has sleep s
118 introduced in Nehalem and SandyBridge processo
119 Thus on an IvyBridge processor one will get Ne
120 state monitors.
121 HaswellExtended extra package sleep state capa
122 specific Haswell (family 0x45) and probably al
123
124 .SS "Fam_12h" "Fam_14h"
125 AMD laptop and desktop processor (family 12h a
126 The registers are accessed via PCI and therefo
127 cores have been offlined.
128
129 There is one special counter: NBP1 (North Brid
130 This one always returns 0 or 1, depending on w
131 power state got entered at least once during m
132 Being able to enter NBP1 state also depends on
133 Therefore this counter can be used to verify w
134 power management is working as expected.
135
136 .SH EXAMPLES
137
138 cpupower monitor -l" may show:
139 .RS 4
140 Monitor "Mperf" (3 states) \- Might overflow a
141
142 ...
143
144 Monitor "Idle_Stats" (3 states) \- Might overf
145
146 ...
147
148 .RE
149 cpupower monitor \-m "Idle_Stats,Mperf" scp /t
150
151 Monitor the scp command, show both Mperf and I
152 statistics, but in exchanged order.
153
154
155
156 .RE
157 Be careful that the typical command to fully u
158
159 cpupower monitor cat /dev/zero >/dev/null
160
161 Does not work as expected, because the measure
162 /dev/null. This could get workarounded by putt
163 shell script. Hit CTRL\-c to terminate the com
164 displayed.
165
166 .SH REFERENCES
167 "BIOS and Kernel Developer’s Guide (BKDG) fo
168 https://support.amd.com/us/Processor_TechDocs/
169
170 "What Is Intel® Turbo Boost Technology?"
171 https://www.intel.com/content/www/us/en/gaming
172
173 "Power Management - Technology Overview"
174 https://cdrdv2.intel.com/v1/dl/getContent/6377
175
176 "Intel® 64 and IA-32 Architectures Software D
177 Volume 3B: System Programming Guide"
178 https://www.intel.com/products/processor/manua
179
180 .SH FILES
181 .ta
182 .nf
183 /dev/cpu/*/msr
184 /sys/devices/system/cpu/cpu*/cpuidle/state*/.
185 .fi
186
187 .SH "SEE ALSO"
188 powertop(8), msr(4), vmstat(8)
189 .PP
190 .SH AUTHORS
191 .nf
192 Written by Thomas Renninger <trenn@suse.de>
193
194 Nehalem, SandyBridge monitors and command pass
195 based on turbostat.8 from Len Brown <len.brown@
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.