1LIBPFM(3) Linux Programmer's Manual LIBPFM(3)
2
3
4
6 libpfm_intel_ivb - support for Intel Ivy Bridge core PMU
7
9 #include <perfmon/pfmlib.h>
10
11 PMU name: ivb
12 PMU desc: Intel Ivy Bridge
13 PMU name: ivb_ep
14 PMU desc: Intel Ivy Bridge EP
15
16
18 The library supports the Intel Ivy Bridge core PMU. It should be noted
19 that this PMU model only covers each core's PMU and not the socket
20 level PMU.
21
22 On Ivy Bridge, the number of generic counters depends on the Hyper‐
23 threading (HT) mode. When HT is on, then only 4 generic counters are
24 available. When HT is off, then 8 generic counters are available. The
25 pfm_get_pmu_info() function returns the maximum number of generic coun‐
26 ters in num_cntrs.
27
28
30 The following modifiers are supported on Intel Ivy Bridge processors:
31
32 u Measure at user level which includes privilege levels 1, 2, 3.
33 This corresponds to PFM_PLM3. This is a boolean modifier.
34
35 k Measure at kernel level which includes privilege level 0. This
36 corresponds to PFM_PLM0. This is a boolean modifier.
37
38 i Invert the meaning of the event. The counter will now count
39 cycles in which the event is not occurring. This is a boolean
40 modifier
41
42 e Enable edge detection, i.e., count only when there is a state
43 transition from no occurrence of the event to at least one
44 occurrence. This modifier must be combined with a counter mask
45 modifier (m) with a value greater or equal to one. This is a
46 boolean modifier.
47
48 c Set the counter mask value. The mask acts as a threshold. The
49 counter will count the number of cycles in which the number of
50 occurrences of the event is greater or equal to the threshold.
51 This is an integer modifier with values in the range [0:255].
52
53 t Measure on both threads at the same time assuming hyper-thread‐
54 ing is enabled. This is a boolean modifier.
55
56 ldlat Pass a latency threshold to the
57 MEM_TRANS_RETIRED:LATENCY_ABOVE_THRESHOLD event. This is an
58 integer attribute that must be in the range [3:65535]. It is
59 required for this event. Note that the event must be used with
60 precise sampling (PEBS).
61
62
64 Intel Ivy Bridge provides two offcore_response events. They are called
65 OFFCORE_RESPONSE_0 and OFFCORE_RESPONSE_1.
66
67 Those events need special treatment in the performance monitoring in‐
68 frastructure because each event uses an extra register to store some
69 settings. Thus, in case multiple offcore_response events are monitored
70 simultaneously, the kernel needs to manage the sharing of that extra
71 register.
72
73 The offcore_response events are exposed as a normal events by the
74 library. The extra settings are exposed as regular umasks. The library
75 takes care of encoding the events according to the underlying kernel
76 interface.
77
78 On Intel Ivy Bridge, the umasks are divided into three categories:
79 request, supplier and snoop. The user must provide at least one umask
80 for each category. The categories are shown in the umask descriptions.
81
82 There is also the special response umask called ANY_RESPONSE. When this
83 umask is used then it overrides any supplier and snoop umasks. In other
84 words, users can specify either ANY_RESPONSE OR any combinations of
85 supplier + snoops.
86
87 In case no supplier or snoop is specified, the library defaults to
88 using ANY_RESPONSE.
89
90 For instance, the following are valid event selections:
91
92 OFFCORE_RESPONSE_0:DMND_DATA_RD:ANY_RESPONSE
93
94 OFFCORE_RESPONSE_0:ANY_REQUEST
95
96 OFFCORE_RESPONSE_0:ANY_RFO:LLC_HITM:SNOOP_ANY
97
98
99 But the following are illegal:
100
101
102 OFFCORE_RESPONSE_0:ANY_RFO:LLC_HITM:ANY_RESPONSE
103
104 OFFCORE_RESPONSE_0:ANY_RFO:LLC_HITM:SNOOP_ANY:ANY_RESPONSE
105
106
108 Stephane Eranian <eranian@gmail.com>
109
110
111
112 August, 2012 LIBPFM(3)