DescOptions

Description of possible counter options

One big novelty in LIKWID 4 are counter options to further filter or extend the incrementing events. Most LIKWID counters can be configured through the config registers but some require the programming of additional registers.

Options without arguments

anythread

This options enables counting the associated event conditions occurring across all logical processors sharing a processor core.

kernel

Specifies that the selected microarchitectural condition is also counted when the logical processor is operating at privilege level 0

edgedetect

Enables (when set) edge detection of the selected microarchitectural condition. The logical processor counts the number of deasserted to asserted transitions for any condition that can be expressed by the other fields. The mechanism does not permit back-to-back assertions to be distinguished.
This mechanism allows software to measure not only the fraction of time spent in a particular state, but also the average length of time spent in such a state (for example, the time spent waiting for an interrupt to be serviced).

invert

When set, inverts the counter-mask (CMASK) comparison, so that both greater than or equal to and less than comparisons can be made (0: greater than or equal; 1: less than). Note if counter-mask is programmed to zero, INV flag is ignored.

occ_edgedetect

Rather than accumulating the raw count each cycle (for events that can increment by 1 per cycle), the register can capture transitions from no event to an event incoming for the PCU’s occupancy events.
Notice: Only for WBOX uncore counters.

occ_invert

Changes the threshold test condition to ‘<‘ for the occupancy events.
Notice: Only for WBOX uncore counters.

in_transaction

When set, the counter will only include counts that occurred inside a transactional region, regardless of whether that region was aborted or committed. This bit may only be set if the processor supports HLE or RTM.
Notice: Only for TSX aware systems (Haswell and newer).

in_transaction_aborted

When set, the counter will not include counts that occurred inside of an aborted transactional region. This bit may only be set if the processor supports HLE or RTM.
Notice: Only for TSX aware systems (Haswell and newer).

Options with arguments

threshold=0xXXXX

When this field is not zero, a logical processor compares this mask to the events count of the detected microarchitectural condition during a single cycle. If the event count is greater than or equal to this mask, the counter is incremented by one. Otherwise the counter is not incremented.
This mask is intended for software to characterize microarchitectural conditions that can count multiple occurrences per cycle (for example, two or more instructions retired per clock; or bus queue occupations). If the counter-mask field is 0, then the counter is incremented each cycle by the event count associated with multiple occurrences.

opcode=0xXXXX

Filter event stream to count only events with matching opcode. Lists of opcodes can be found in Intel® Software Developer System Programming Manual, Vol. 3, Chapter Performance Monitoring or the Intel Uncore manuals.

nid=0xXXXX

Filter event stream to count only events with matching NUMA node ID. The argument is a bitmask representing the NUMA node IDs, consequently NUMA node 0 has the bitmask 0x1

tid=0xXXXX

Filter event stream to count only events with matching thread ID. For the structure of this field check the vendor documentation. Often, the highest bit is used to select the physical hardware thread or the HyperThreading virtual thread while the lower bits define the CPU core.

state=0xXXXX

Filter event stream to count only events with matching state, e.g. cache line state.

occupancy=0xXXXX

Filter event stream to count only events when event target is occupied with argument.
For example, if a queue is known to contain eight entries, it may be useful to know how often it contains 6 or more entires (i.e. Almost Full) or when it contains 1 or more entries (i.e. Not Empty).

occupancy_filter=0xXXXX

Filter event stream to count only events when event target is occupied and fit the given filter.
Notice: Only for WBOX uncore counters.

match<0-3>=0xXXXX

Generic options used mainly for setting external filter registers to match something specific. For example, the options match0 and match1 are used to filter offcore events for the event OFFCORE_RESPONSE_0_OPTIONS, where match0 sets the request part and match1 the response part of the OFFCORE_RESPONSE register.

mask<0-3>=0xXXXX

Generic options used mainly for setting external filter registers to mask something specific. Often used in combination with the match<0-3> events.

cid=0xXXXX

Option used to select core-id.

slice=0xXXXX

Option used to select L3 slide