1LIBPFM(3) Linux Programmer's Manual LIBPFM(3)
2
6 libpfm_intel_nhm - support for Intel Nehalem core PMU
7
9 #include <perfmon/pfmlib.h>
10 PMU name: nhm
12 PMU desc: Intel Nehalem
13 PMU name: nhm_ex
14 PMU desc: Intel Nehalem EX
15
18 The library supports the Intel Nehalem core PMU. It should be noted
19 that this PMU model only covers the each core's PMU and not the socket
20 level PMU. It is provided separately. Support is provided for the Intel
21 Core i7 and Core i5 processors.
22
25 The following modifiers are supported on Intel Nehalem processors:
26 u Measure at user level which includes privilege levels 1, 2, 3.
28 This corresponds to PFM_PLM3. This is a boolean modifier.
29 k Measure at kernel level which includes privilege level 0. This
31 corresponds to PFM_PLM0. This is a boolean modifier.
32 i Invert the meaning of the event. The counter will now count cy‐
34 cles in which the event is not occurring. This is a boolean mod‐
35 ifier
36 e Enable edge detection, i.e., count only when there is a state
38 transition from no occurrence of the event to at least one oc‐
39 currence. This modifier must be combined with a counter mask
40 modifier (m) with a value greater or equal to one. This is a
41 boolean modifier.
42 c Set the counter mask value. The mask acts as a threshold. The
44 counter will count the number of cycles in which the number of
45 occurrences of the event is greater or equal to the threshold.
46 This is an integer modifier with values in the range [0:255].
47 t Measure on both threads at the same time assuming hyper-thread‐
49 ing is enabled. This is a boolean modifier.
50 ldlat Pass a latency threshold to the MEM_INST_RETIRED:LA‐
52 TENCY_ABOVE_THRESHOLD event. This is an integer attribute that
53 must be in the range [1:65535]. It is required for this event.
54 Note that the event must be used with precise sampling (PEBS).
55
58 The library is able to encode the OFFCORE_RESPONSE_0 event. This is a
59 special event because it needs a second MSR (0x1a6) to be programmed
60 for the event to count properly. Thus two values are necessary. The
61 first value can be programmed on any of the generic counters. The sec‐
62 ond value goes into the dedicated MSR (0x1a6).
63 The OFFCORE_RESPONSE event is exposed as a normal event with several
65 umasks which are divided in two groups: request and response. The user
66 must provide at least one umask from each group. For instance, OFF‐
67 CORE_RESPONSE_0:ANY_DATA:LOCAL_DRAM.
68 When using pfm_get_event_encoding(), two 64-bit values are returned.
70 The first value corresponds to what needs to be programmed into any of
71 the generic counters. The second value must be programmed into the ded‐
72 icated MSR (0x1a6).
73 When using an OS-specific encoding routine, the way the event is en‐
75 coded is OS specific. Refer to the corresponding man page for more in‐
76 formation.
77
80 Stephane Eranian <eranian@gmail.com>
81 September, 2009 LIBPFM(3)