Basic core counter state. More...

#include <cpucounters.h>

Inheritance diagram for pcm::BasicCounterState:

Collaboration diagram for pcm::BasicCounterState:

Public Member Functions
	BasicCounterState (const BasicCounterState &)=default

	BasicCounterState (BasicCounterState &&)=default

BasicCounterState &	operator= (BasicCounterState &&)=default

BasicCounterState &	operator+= (const BasicCounterState &o)

void	readAndAggregate (std::shared_ptr< SafeMsrHandle >)

void	readAndAggregateTSC (std::shared_ptr< SafeMsrHandle >)

int32	getThermalHeadroom () const
	Returns current thermal headroom below TjMax.

Protected Types
enum	{ L3MissPos = 0, ArchLLCMissPos = 0, L3UnsharedHitPos = 1, ArchLLCRefPos = 1, SKLL3HitPos = 1, L2HitMPos = 2, SKLL2MissPos = 2, L2HitPos = 3 }

Protected Attributes
checked_uint64	InstRetiredAny {}

checked_uint64	CpuClkUnhaltedThread {}

checked_uint64	CpuClkUnhaltedRef {}

checked_uint64	Event [PERF_MAX_CUSTOM_COUNTERS]

uint64	InvariantTSC

uint64	CStateResidency [PCM::MAX_C_STATE+1]

int32	ThermalHeadroom

uint64	L3Occupancy

uint64	MemoryBWLocal

uint64	MemoryBWTotal

uint64	SMICount

uint64	FrontendBoundSlots

uint64	BadSpeculationSlots

uint64	BackendBoundSlots

uint64	RetiringSlots

uint64	AllSlotsRaw

std::unordered_map< uint64, uint64 >	MSRValues

Friends
class	PCM

class	JSONPrinter

template<class CounterStateType >
double	getExecUsage (const CounterStateType &before, const CounterStateType &after)
	Computes average number of retired instructions per time interval. More...

template<class CounterStateType >
double	getIPC (const CounterStateType &before, const CounterStateType &after)
	Computes average number of retired instructions per core cycle (IPC) More...

template<class CounterStateType >
double	getAverageFrequency (const CounterStateType &before, const CounterStateType &after)
	Computes average core frequency also taking Intel Turbo Boost technology into account. More...

template<class CounterStateType >
double	getAverageFrequencyFromClocks (const int64 clocks, const CounterStateType &before, const CounterStateType &after)

template<class CounterStateType >
double	getActiveAverageFrequency (const CounterStateType &before, const CounterStateType &after)
	Computes average core frequency when not in powersaving C0-state (also taking Intel Turbo Boost technology into account) More...

template<class CounterStateType >
double	getRelativeFrequency (const CounterStateType &before, const CounterStateType &after)
	Computes average core frequency also taking Intel Turbo Boost technology into account. More...

template<class CounterStateType >
double	getActiveRelativeFrequency (const CounterStateType &before, const CounterStateType &after)
	Computes average core frequency when not in powersaving C0-state (also taking Intel Turbo Boost technology into account) More...

template<class CounterStateType >
double	getL2CacheHitRatio (const CounterStateType &before, const CounterStateType &after)
	Computes L2 cache hit ratio. More...

template<class CounterStateType >
double	getL3CacheHitRatio (const CounterStateType &before, const CounterStateType &after)
	Computes L3 cache hit ratio. More...

template<class CounterStateType >
uint64	getL3CacheMisses (const CounterStateType &before, const CounterStateType &after)
	Computes number of L3 cache misses. More...

template<class CounterStateType >
uint64	getL2CacheMisses (const CounterStateType &before, const CounterStateType &after)
	Computes number of L2 cache misses. More...

template<class CounterStateType >
uint64	getL2CacheHits (const CounterStateType &before, const CounterStateType &after)
	Computes number of L2 cache hits. More...

template<class CounterStateType >
uint64	getL3CacheHitsNoSnoop (const CounterStateType &before, const CounterStateType &after)
	Computes number of L3 cache hits where no snooping in sibling L2 caches had to be done. More...

template<class CounterStateType >
uint64	getL3CacheHitsSnoop (const CounterStateType &before, const CounterStateType &after)
	Computes number of L3 cache hits where snooping in sibling L2 caches had to be done. More...

template<class CounterStateType >
uint64	getL3CacheHits (const CounterStateType &before, const CounterStateType &after)
	Computes total number of L3 cache hits. More...

template<class CounterStateType >
uint64	getL3CacheOccupancy (const CounterStateType &now)
	Computes L3 Cache Occupancy. More...

template<class CounterStateType >
uint64	getLocalMemoryBW (const CounterStateType &before, const CounterStateType &after)
	Computes Local Memory Bandwidth. More...

template<class CounterStateType >
uint64	getRemoteMemoryBW (const CounterStateType &before, const CounterStateType &after)
	Computes Remote Memory Bandwidth. More...

template<class CounterStateType >
uint64	getCycles (const CounterStateType &before, const CounterStateType &after)
	Computes the number core clock cycles when signal on a specific core is running (not halted) More...

template<class CounterStateType >
uint64	getInstructionsRetired (const CounterStateType &before, const CounterStateType &after)
	Computes the number of retired instructions. More...

template<class CounterStateType >
uint64	getCycles (const CounterStateType &now)
	Computes the number executed core clock cycles. More...

template<class CounterStateType >
uint64	getInstructionsRetired (const CounterStateType &now)
	Computes the number of retired instructions. More...

template<class CounterStateType >
uint64	getNumberOfCustomEvents (int32 eventCounterNr, const CounterStateType &before, const CounterStateType &after)
	Returns the number of occurred custom core events. More...

template<class CounterStateType >
uint64	getInvariantTSC (const CounterStateType &before, const CounterStateType &after)
	Computes number of invariant time stamp counter ticks. More...

template<class CounterStateType >
uint64	getRefCycles (const CounterStateType &before, const CounterStateType &after)
	Computes the number of reference clock cycles while clock signal on the core is running. More...

template<class CounterStateType >
double	getCoreCStateResidency (int state, const CounterStateType &before, const CounterStateType &after)
	Computes residency in the core C-state. More...

template<class CounterStateType >
uint64	getCoreCStateResidency (int state, const CounterStateType &now)
	Reads raw residency counter for the core C-state. More...

template<class CounterStateType >
uint64	getSMICount (const CounterStateType &before, const CounterStateType &after)
	Returns the number of occurred system management interrupts. More...

template<class CounterStateType >
uint64	getAllSlotsRaw (const CounterStateType &before, const CounterStateType &after)

template<class CounterStateType >
uint64	getAllSlots (const CounterStateType &before, const CounterStateType &after)

template<class CounterStateType >
double	getBackendBound (const CounterStateType &before, const CounterStateType &after)
	Returns unutilized pipeline slots where no uop was delivered due to lack of back-end resources as range 0..1.

template<class CounterStateType >
double	getFrontendBound (const CounterStateType &before, const CounterStateType &after)
	Returns unutilized pipeline slots where Front-end did not deliver a uop while back-end is ready as range 0..1.

template<class CounterStateType >
double	getBadSpeculation (const CounterStateType &before, const CounterStateType &after)
	Returns wasted pipeline slots due to incorrect speculation, covering whole penalty: Utilized by uops that do not retire, or Recovery Bubbles (unutilized slots) as range 0..1.

template<class CounterStateType >
double	getRetiring (const CounterStateType &before, const CounterStateType &after)
	Returns pipeline slots utilized by uops that eventually retire (commit)

template<class CounterStateType >
uint64	getMSREvent (const uint64 &index, const PCM::MSRType &type, const CounterStateType &before, const CounterStateType &after)

Detailed Description

Basic core counter state.

Intended only for derivation, but not for the direct use

Friends And Related Function Documentation

◆ getActiveAverageFrequency

template<class CounterStateType >

double getActiveAverageFrequency	(	const CounterStateType &	before,
		const CounterStateType &	after
	)

friend

Computes average core frequency when not in powersaving C0-state (also taking Intel Turbo Boost technology into account)

Parameters

before	CPU counter state before the experiment
after	CPU counter state after the experiment

Returns: frequency in Hz

◆ getActiveRelativeFrequency

template<class CounterStateType >

double getActiveRelativeFrequency	(	const CounterStateType &	before,
		const CounterStateType &	after
	)

friend

Computes average core frequency when not in powersaving C0-state (also taking Intel Turbo Boost technology into account)

Parameters

before	CPU counter state before the experiment
after	CPU counter state after the experiment

Returns: Fraction of nominal frequency (if >1.0 then Turbo was working during the measurement)

◆ getAverageFrequency

template<class CounterStateType >

double getAverageFrequency	(	const CounterStateType &	before,
		const CounterStateType &	after
	)

friend

Computes average core frequency also taking Intel Turbo Boost technology into account.

Parameters

before	CPU counter state before the experiment
after	CPU counter state after the experiment

Returns: frequency in Hz

◆ getCoreCStateResidency [1/2]

template<class CounterStateType >

double getCoreCStateResidency	(	int	state,
		const CounterStateType &	before,
		const CounterStateType &	after
	)

friend

Computes residency in the core C-state.

Parameters

state	C-state
before	CPU counter state before the experiment
after	CPU counter state after the experiment

Returns: residence ratio (0..1): 0 - 0%, 1.0 - 100%

◆ getCoreCStateResidency [2/2]

template<class CounterStateType >

uint64 getCoreCStateResidency	(	int	state,
		const CounterStateType &	now
	)

friend

Reads raw residency counter for the core C-state.

Parameters

state	C-state #
now	CPU counter state

Returns: raw residency value

◆ getCycles [1/2]

template<class CounterStateType >

uint64 getCycles	(	const CounterStateType &	before,
		const CounterStateType &	after
	)

friend

Computes the number core clock cycles when signal on a specific core is running (not halted)

Returns number of used cycles (halted cyles are not counted). The counter does not advance in the following conditions:

an ACPI C-state is other than C0 for normal operation
HLT
STPCLK+ pin is asserted
being throttled by TM1
during the frequency switching phase of a performance state transition

The performance counter for this event counts across performance state transitions using different core clock frequencies

Parameters

before	CPU counter state before the experiment
after	CPU counter state after the experiment

Returns: number core clock cycles

◆ getCycles [2/2]

template<class CounterStateType >

uint64 getCycles ( const CounterStateType & now )

friend

Computes the number executed core clock cycles.

Returns number of used cycles (halted cyles are not counted).

Parameters

now	Current CPU counter state

Returns: number core clock cycles

◆ getExecUsage

template<class CounterStateType >

double getExecUsage	(	const CounterStateType &	before,
		const CounterStateType &	after
	)

friend

Computes average number of retired instructions per time interval.

Parameters

before	CPU counter state before the experiment
after	CPU counter state after the experiment

Returns: usage

◆ getInstructionsRetired [1/2]

template<class CounterStateType >

uint64 getInstructionsRetired	(	const CounterStateType &	before,
		const CounterStateType &	after
	)

friend

Computes the number of retired instructions.

Parameters

before	CPU counter state before the experiment
after	CPU counter state after the experiment

Returns: number of retired instructions

◆ getInstructionsRetired [2/2]

template<class CounterStateType >

uint64 getInstructionsRetired ( const CounterStateType & now )

friend

Computes the number of retired instructions.

Parameters

now	Current CPU counter state

Returns: number of retired instructions

◆ getInvariantTSC

template<class CounterStateType >

uint64 getInvariantTSC	(	const CounterStateType &	before,
		const CounterStateType &	after
	)

friend

Computes number of invariant time stamp counter ticks.

This counter counts irrespectively of C-, P- or T-states

Parameters

before	CPU counter state before the experiment
after	CPU counter state after the experiment

Returns: number of time stamp counter ticks

◆ getIPC

template<class CounterStateType >

double getIPC	(	const CounterStateType &	before,
		const CounterStateType &	after
	)

friend

Computes average number of retired instructions per core cycle (IPC)

Parameters

before	CPU counter state before the experiment
after	CPU counter state after the experiment

Returns: IPC

◆ getL2CacheHitRatio

template<class CounterStateType >

double getL2CacheHitRatio	(	const CounterStateType &	before,
		const CounterStateType &	after
	)

friend

Computes L2 cache hit ratio.

Parameters

before	CPU counter state before the experiment
after	CPU counter state after the experiment

Warning: Works only in the DEFAULT_EVENTS programming mode (see program() method)

Returns: value between 0 and 1

◆ getL2CacheHits

template<class CounterStateType >

uint64 getL2CacheHits	(	const CounterStateType &	before,
		const CounterStateType &	after
	)

friend

Computes number of L2 cache hits.

Parameters

before	CPU counter state before the experiment
after	CPU counter state after the experiment

Warning: Works only in the DEFAULT_EVENTS programming mode (see program() method)

Returns: number of hits

◆ getL2CacheMisses

template<class CounterStateType >

uint64 getL2CacheMisses	(	const CounterStateType &	before,
		const CounterStateType &	after
	)

friend

Computes number of L2 cache misses.

Parameters

before	CPU counter state before the experiment
after	CPU counter state after the experiment

Warning: Works only in the DEFAULT_EVENTS programming mode (see program() method)

Returns: number of misses

◆ getL3CacheHitRatio

template<class CounterStateType >

double getL3CacheHitRatio	(	const CounterStateType &	before,
		const CounterStateType &	after
	)

friend

Computes L3 cache hit ratio.

Parameters

before	CPU counter state before the experiment
after	CPU counter state after the experiment

Warning: Works only in the DEFAULT_EVENTS programming mode (see program() method)

Returns: value between 0 and 1

◆ getL3CacheHits

template<class CounterStateType >

uint64 getL3CacheHits	(	const CounterStateType &	before,
		const CounterStateType &	after
	)

friend

Computes total number of L3 cache hits.

Parameters

before	CPU counter state before the experiment
after	CPU counter state after the experiment

Warning: Works only in the DEFAULT_EVENTS programming mode (see program() method)

Returns: number of hits

◆ getL3CacheHitsNoSnoop

template<class CounterStateType >

uint64 getL3CacheHitsNoSnoop	(	const CounterStateType &	before,
		const CounterStateType &	after
	)

friend

Computes number of L3 cache hits where no snooping in sibling L2 caches had to be done.

Parameters

before	CPU counter state before the experiment
after	CPU counter state after the experiment

Warning: Works only in the DEFAULT_EVENTS programming mode (see program() method)

Returns: number of hits

◆ getL3CacheHitsSnoop

template<class CounterStateType >

uint64 getL3CacheHitsSnoop	(	const CounterStateType &	before,
		const CounterStateType &	after
	)

friend

Computes number of L3 cache hits where snooping in sibling L2 caches had to be done.

Parameters

before	CPU counter state before the experiment
after	CPU counter state after the experiment

Warning: Works only in the DEFAULT_EVENTS programming mode (see program() method)

Returns: number of hits

◆ getL3CacheMisses

template<class CounterStateType >

uint64 getL3CacheMisses	(	const CounterStateType &	before,
		const CounterStateType &	after
	)

friend

Computes number of L3 cache misses.

Parameters

before	CPU counter state before the experiment
after	CPU counter state after the experiment

Warning: Works only in the DEFAULT_EVENTS programming mode (see program() method)

Returns: number of misses

◆ getL3CacheOccupancy

template<class CounterStateType >

uint64 getL3CacheOccupancy ( const CounterStateType & now )

friend

Computes L3 Cache Occupancy.

◆ getLocalMemoryBW

template<class CounterStateType >

uint64 getLocalMemoryBW	(	const CounterStateType &	before,
		const CounterStateType &	after
	)

friend

Computes Local Memory Bandwidth.

◆ getNumberOfCustomEvents

template<class CounterStateType >

uint64 getNumberOfCustomEvents	(	int32	eventCounterNr,
		const CounterStateType &	before,
		const CounterStateType &	after
	)

friend

Returns the number of occurred custom core events.

Read number of events programmed with the CUSTOM_CORE_EVENTS

Parameters

eventCounterNr	Event/counter number (value from 0 to 3)
before	CPU counter state before the experiment
after	CPU counter state after the experiment

Returns: Number of bytes

◆ getRefCycles

template<class CounterStateType >

uint64 getRefCycles	(	const CounterStateType &	before,
		const CounterStateType &	after
	)

friend

Computes the number of reference clock cycles while clock signal on the core is running.

The reference clock operates at a fixed frequency, irrespective of core frequency changes due to performance state transitions. See Intel(r) Software Developer's Manual for more details

Parameters

before	CPU counter state before the experiment
after	CPU counter state after the experiment

Returns: number core clock cycles

◆ getRelativeFrequency

template<class CounterStateType >

double getRelativeFrequency	(	const CounterStateType &	before,
		const CounterStateType &	after
	)

friend

Computes average core frequency also taking Intel Turbo Boost technology into account.

Parameters

before	CPU counter state before the experiment
after	CPU counter state after the experiment

Returns: Fraction of nominal frequency

◆ getRemoteMemoryBW

template<class CounterStateType >

uint64 getRemoteMemoryBW	(	const CounterStateType &	before,
		const CounterStateType &	after
	)

friend

Computes Remote Memory Bandwidth.

◆ getSMICount

template<class CounterStateType >

uint64 getSMICount	(	const CounterStateType &	before,
		const CounterStateType &	after
	)

friend

Returns the number of occurred system management interrupts.

Parameters

before	CPU counter state before the experiment
after	CPU counter state after the experiment

Returns: Number of SMIs (system manegement interrupts)

The documentation for this class was generated from the following files:

src/cpucounters.h
src/cpucounters.cpp

Public Member Functions

Protected Types

Protected Attributes

Friends

Detailed Description

Friends And Related Function Documentation

◆ getActiveAverageFrequency

◆ getActiveRelativeFrequency

◆ getAverageFrequency

◆ getCoreCStateResidency [1/2]

◆ getCoreCStateResidency [2/2]

◆ getCycles [1/2]

◆ getCycles [2/2]

◆ getExecUsage

◆ getInstructionsRetired [1/2]

◆ getInstructionsRetired [2/2]

◆ getInvariantTSC

◆ getIPC

◆ getL2CacheHitRatio

◆ getL2CacheHits

◆ getL2CacheMisses

◆ getL3CacheHitRatio

◆ getL3CacheHits

◆ getL3CacheHitsNoSnoop

◆ getL3CacheHitsSnoop

◆ getL3CacheMisses

◆ getL3CacheOccupancy

◆ getLocalMemoryBW

◆ getNumberOfCustomEvents

◆ getRefCycles

◆ getRelativeFrequency

◆ getRemoteMemoryBW

◆ getSMICount