This commit was generated by cvs2svn to compensate for changes in r177580,

[FreeBSD/FreeBSD.git] / lib / libpmc / pmc.3

.\" Copyright (c) 2003-2008 Joseph Koshy.  All rights reserved.
.\"
.\" Redistribution and use in source and binary forms, with or without
.\" modification, are permitted provided that the following conditions
.\" are met:
.\" 1. Redistributions of source code must retain the above copyright
.\"    notice, this list of conditions and the following disclaimer.
.\" 2. Redistributions in binary form must reproduce the above copyright
.\"    notice, this list of conditions and the following disclaimer in the
.\"    documentation and/or other materials provided with the distribution.
.\"
.\" This software is provided by Joseph Koshy ``as is'' and
.\" any express or implied warranties, including, but not limited to, the
.\" implied warranties of merchantability and fitness for a particular purpose
.\" are disclaimed.  in no event shall Joseph Koshy be liable
.\" for any direct, indirect, incidental, special, exemplary, or consequential
.\" damages (including, but not limited to, procurement of substitute goods
.\" or services; loss of use, data, or profits; or business interruption)
.\" however caused and on any theory of liability, whether in contract, strict
.\" liability, or tort (including negligence or otherwise) arising in any way
.\" out of the use of this software, even if advised of the possibility of
.\" such damage.
.\"
.\" $FreeBSD$
.\"
.Dd March 14, 2008
.Os
.Dt PMC 3
.Sh NAME
.Nm pmc
.Nd library for accessing hardware performance monitoring counters
.Sh LIBRARY
.Lb libpmc
.Sh SYNOPSIS
.In pmc.h
.Sh DESCRIPTION
The
.Lb libpmc
provides a programming interface that allows applications to use
hardware performance counters to gather performance data about
specific processes or for the system as a whole.
The library is implemented using the lower-level facilities offered by
the
.Xr hwpmc 4
driver.
.Ss Key Concepts
Performance monitoring counters (PMCs) are represented by the library
using a software abstraction.
These
.Dq abstract
PMCs can have one two scopes:
.Bl -bullet
.It
System scope.
These PMCs measure events in a whole-system manner, i.e., independent
of the currently executing thread.
System scope PMCs are allocated on specific CPUs and do not
migrate between CPUs.
Non-privileged process are allowed to allocate system scope PMCs if the
.Xr hwpmc 4
sysctl tunable:
.Va security.bsd.unprivileged_syspmcs
is non-zero.
.It
Process scope.
These PMCs only measure hardware events when the processes they are
attached to are executing on a CPU.
In an SMP system, process scope PMCs migrate between CPUs along with
their target processes.
.El
.Pp
Orthogonal to PMC scope, PMCs may be allocated in one of two
operational modes:
.Bl -bullet
.It
Counting PMCs measure events according to their scope
(system or process).
The application needs to explicitly read these counters
to retrieve their value.
.It
Sampling PMCs cause the CPU to be periodically interrupted
and information about its state of execution to be collected.
Sampling PMCs are used to profile specific processes and kernel
threads or to profile the system as a whole.
.El
.Pp
The scope and operational mode for a software PMC are specified at
PMC allocation time.
An application is allowed to allocate multiple PMCs subject
to availability of hardware resources.
.Pp
The library uses human-readable strings to name the event being
measured by hardware.
The syntax used for specifying a hardware event along with additional
event specific qualifiers (if any) is described in detail in section
.Sx "EVENT SPECIFIERS"
below.
.Pp
PMCs are associated with the process that allocated them and
will be automatically reclaimed by the system when the process exits.
Additionally, process-scope PMCs have to be attached to one or more
target processes before they can perform measurements.
A process-scope PMC may be attached to those target processes
that its owner process would otherwise be permitted to debug.
An owner process may attach PMCs to itself allowing
it to measure its own behavior.
Additionally, on some machine architectures, such self-attached PMCs
may be read cheaply using specialized instructions supported by the
processor.
.Pp
Certain kinds of PMCs require that a log file be configured before
they may be started.
These include:
.Bl -bullet -compact
.It
System scope sampling PMCs.
.It
Process scope sampling PMCs.
.It
Process scope counting PMCs that have been configured to report PMC
readings on process context switches or process exits.
.El
Upto one log file may be configured per owner process.
Events logged to a log file may be subsequently analyzed using the
.Xr pmclog 3
family of functions.
.Ss Supported CPUs
The CPUs known to the PMC library are named by the
.Vt "enum pmc_cputype"
enumeration.
Supported CPUs include:
.Bl -tag -width PMC_CPU_INTEL_PIII -compact
.It PMC_CPU_AMD_K7
.Tn "AMD Athlon"
CPUs.
.It PMC_CPU_AMD_K8
.Tn "AMD Athlon64"
CPUs.
.It PMC_CPU_INTEL_P5
.Tn Intel
.Tn "Pentium"
CPUs.
.It PMC_CPU_INTEL_P6
.Tn Intel
.Tn "Pentium Pro"
CPUs.
.It PMC_CPU_INTEL_PII
.Tn "Intel Pentium II"
CPUs.
.It PMC_CPU_INTEL_PIII
.Tn "Intel Pentium III"
CPUs.
.It PMC_CPU_INTEL_PM
.Tn "Intel Pentium M"
CPUs.
.It PMC_CPU_INTEL_PIV
.Tn "Intel Pentium 4"
CPUs.
.El
.Ss Supported PMCs
PMC supported by this library are named by the
.Vt enum pmc_class
enumeration.
Supported PMC kinds include:
.Bl -tag -width PMC_CLASS_TSC -compact
.It PMC_CLASS_TSC
The timestamp counter on i386 and amd64 architecture CPUs.
.It PMC_CLASS_K7
Programmable hardware counters present in
.Tn "AMD Athlon"
CPUs.
.It PMC_CLASS_K8
Programmable hardware counters present in
.Tn "AMD Athlon64"
CPUs.
.It PMC_CLASS_P5
Programmable hardware counters present in
.Tn Intel
.Tn Pentium
CPUs.
.It PMC_CLASS_P6
Programmable hardware counters present in
.Tn Intel
.Tn "Pentium Pro" ,
.Tn "Pentium II" ,
.Tn "Pentium III" ,
.Tn "Celeron" ,
and
.Tn "Pentium M"
CPUs.
.It PMC_CLASS_P4
Programmable hardware counters present in
.Tn "Intel Pentium 4"
CPUs.
.El
.Ss PMC Capabilities
.Pp
Capabilities of performance monitoring hardware are denoted using
the
.Vt "enum pmc_caps"
enumeration.
Supported capabilities include:
.Bl -tag -width "PMC_CAP_INTERRUPT" -compact
.It PMC_CAP_EDGE
The ability to count negated to asserted transitions of the hardware
conditions being probed for.
.It PMC_CAP_INTERRUPT
The ability to interrupt the CPU.
.It PMC_CAP_INVERT
The ability to invert the sense of the hardware conditions being
measured.
.It PMC_CAP_READ
PMC hardware allows the CPU to read performance counters.
.It PMC_CAP_QUALIFIER
The hardware allows monitored to be further qualified in some
system dependent way.
.It PMC_CAP_SYSTEM
The ability to restrict counting of hardware events to when the CPU is
running privileged code.
.It PMC_CAP_THRESHOLD
The ability to ignore simultaneous hardware events below a
programmable threshold.
.It PMC_CAP_USER
The ability to restrict counting of hardware events to those when the
CPU is running unprivileged code.
.It PMC_CAP_WRITE
PMC hardware allows CPUs write to counters.
.El
.Ss Functional Grouping
This section contains a brief overview of the available functionality
in the PMC library.
Each function listed here is described further in its own manual page.
.Bl -tag -width indent
.It Administration
.Bl -tag -compact
.It Fn pmc_disable , Fn pmc_enable
Administratively disable (enable) specific performance monitoring
counter hardware.
Counters that are disabled will not be available to applications to
use.
.El
.It "Convenience Functions"
.Bl -tag -compact
.It Fn pmc_event_names_of_class
Returns a list of event names supported by a given PMC type.
.It Fn pmc_name_of_capability
Convert a
.Dv PMC_CAP_*
flag to a human-readable string.
.It Fn pmc_name_of_class
Convert a
.Dv PMC_CLASS_*
constant to a human-readable string.
.It Fn pmc_name_of_cputype
Return a human-readable name for a CPU type.
.It Fn pmc_name_of_disposition
Return a human-readable string describing a PMC's disposition.
.It Fn pmc_name_of_event
Convert a numeric event code to a human-readable string.
.It Fn pmc_name_of_mode
Convert a
.Dv PMC_MODE_*
constant to a human-readable name.
.It Fn pmc_name_of_state
Return a human-readable string describing a PMC's current state.
.El
.It "Library Initialization"
.Bl -tag -compact
.It Fn pmc_init
Initialize the library.
This function must be called before any other library function.
.El
.It "Log File Handling"
.Bl -tag -compact
.It Fn pmc_configure_logfile
Configure a log file for
.Xr hwpmc 4
to write logged events to.
.It Fn pmc_flush_logfile
Flush all pending log data in
.Xr hwpmc 4 Ns Ap s
buffers.
.It Fn pmc_writelog
Append arbitrary user data to the current log file.
.El
.It "PMC Management"
.Bl -tag -compact
.It Fn pmc_allocate , Fn pmc_release
Allocate (free) a PMC.
.It Fn pmc_attach , Fn pmc_detach
Attach (detach) a process scope PMC to a target.
.It Fn pmc_read , Fn pmc_write , Fn pmc_rw
Read (write) a value from (to) a PMC.
.It Fn pmc_start , Fn pmc_stop
Start (stop) a software PMC.
.It Fn pmc_set
Set the reload value for a sampling PMC.
.El
.It "Queries"
.Bl -tag -compact
.It Fn pmc_capabilities
Retrieve the capabilities for a given PMC.
.It Fn pmc_cpuinfo
Retrieve information about the CPUs and PMC hardware present in the
system.
.It Fn pmc_get_driver_stats
Retrieve statistics maintained by
.Xr hwpmc 4 .
.It Fn pmc_ncpu
Determine the number of CPUs in the system.
.It Fn pmc_npmc
Return the number of hardware PMCs present in a given CPU.
.It Fn pmc_pmcinfo
Return information about the state of a given CPU's PMCs.
.It Fn pmc_width
Determine the width of a hardware counter in bits.
.El
.It "x86 Architecture Specific API"
.Bl -tag -compact
.It Fn pmc_get_msr
Returns the processor model specific register number
associated with
.Fa pmc .
Applications may then use the x86
.Ic RDPMC
instruction to directly read the contents of the PMC.
.El
.El
.Ss Signal Handling Requirements
Applications using PMCs are required to handle the following signals:
.Bl -tag -width ".Dv SIGBUS"
.It Dv SIGBUS
When the
.Xr hwpmc 4
module is unloaded using
.Xr kldunload 8 ,
processes that have PMCs allocated to them will be sent a
.Dv SIGBUS
signal.
.It Dv SIGIO
The
.Xr hwpmc 4
driver will send a PMC owning process a
.Dv SIGIO
signal if:
.Bl -bullet
.It
If any process-mode PMC allocated by it loses all its
target processes.
.It
If the driver encounters an error when writing log data to a
configured log file.
This error may be retrieved by a subsequent call to
.Fn pmc_flush_logfile .
.El
.El
.Ss Typical Program Flow
.Bl -enum
.It
An application would first invoke function
.Fn pmc_init
to allow the library to initialize itself.
.It
Signal handling would then be set up.
.It
Next the application would allocate the PMCs it desires using function
.Fn pmc_allocate .
.It
Initial values for PMCs may be set using function
.Fn pmc_set .
.It
If a log file is necessary for the PMCs to work, it would
be configured using function
.Fn pmc_configure_logfile .
.It
Process scope PMCs would then be attached to their target processes
using function
.Fn pmc_attach .
.It
The PMCs would then be started using function
.Fn pmc_start .
.It
Once started, the values of counting PMCs may be read using function
.Fn pmc_start .
For PMCs that write events to the log file, this logged data would be
read and parsed using the
.Xr pmclog 3
family of functions.
.It
PMCs are stopped using function
.Fn pmc_stop ,
and process scope PMCs are detached from their targets using
function
.Fn pmc_detach .
.It
Before the process exits, its may release its PMCs using function
.Fn pmc_release .
Any configured log file may be closed using function
.Fn pmc_configure_logfile .
.El
.Sh EVENT SPECIFIERS
Event specifiers are strings comprising of an event name, followed by
optional parameters modifying the semantics of the hardware event
being probed.
Event names are PMC architecture dependent, but the PMC library defines
machine independent aliases for commonly used events.
.Ss Event Name Aliases
Event name aliases are CPU architecture independent names for commonly
used events.
The following aliases are known to this version of the
.Nm pmc
library:
.Bl -tag -width indent
.It Li branches
Measure the number of branches retired.
.It Li branch-mispredicts
Measure the number of retired branches that were mispredicted.
.It Li cycles
Measure processor cycles.
This event is implemented using the processor's Time Stamp Counter
register.
.It Li dc-misses
Measure the number of data cache misses.
.It Li ic-misses
Measure the number of instruction cache misses.
.It Li instructions
Measure the number of instructions retired.
.It Li interrupts
Measure the number of interrupts seen.
.It Li unhalted-cycles
Measure the number of cycles the processor is not in a halted
or sleep state.
.El
.Ss Time Stamp Counter (TSC)
The timestamp counter is a monotonically non-decreasing counter that
counts processor cycles.
.Pp
In the i386 architecture, this counter may
be selected by requesting an event with event specifier
.Dq Li tsc .
The
.Dq Li tsc
event does not support any further qualifiers.
It can only be allocated in system-wide counting mode,
and is a read-only counter.
Multiple processes are allowed to allocate the TSC.
Once allocated, it may be read using the
.Fn pmc_read
function, or by using the RDTSC instruction.
.Ss AMD (K7) PMCs
These PMCs are present in the
.Tn "AMD Athlon"
series of CPUs and are documented in:
.Rs
.%B "AMD Athlon Processor x86 Code Optimization Guide"
.%N "Publication No. 22007"
.%D "February 2002"
.%Q "Advanced Micro Devices, Inc."
.Re
.Pp
Event specifiers for AMD K7 PMCs can have the following optional
qualifiers:
.Bl -tag -width indent
.It Li count= Ns Ar value
Configure the counter to increment only if the number of configured
events measured in a cycle is greater than or equal to
.Ar value .
.It Li edge
Configure the counter to only count negated-to-asserted transitions
of the conditions expressed by the other qualifiers.
In other words, the counter will increment only once whenever a given
condition becomes true, irrespective of the number of clocks during
which the condition remains true.
.It Li inv
Invert the sense of comparision when the
.Dq Li count
qualifier is present, making the counter to increment when the
number of events per cycle is less than the value specified by
the
.Dq Li count
qualifier.
.It Li os
Configure the PMC to count events happening at privilege level 0.
.It Li unitmask= Ns Ar mask
This qualifier is used to further qualify a select few events,
.Dq Li k7-dc-refills-from-l2 ,
.Dq Li k7-dc-refills-from-system
and
.Dq Li k7-dc-writebacks .
Here
.Ar mask
is a string of the following characters optionally separated by
.Ql +
characters:
.Pp
.Bl -tag -width indent -compact
.It Li m
Count operations for lines in the
.Dq Modified
state.
.It Li o
Count operations for lines in the
.Dq Owner
state.
.It Li e
Count operations for lines in the
.Dq Exclusive
state.
.It Li s
Count operations for lines in the
.Dq Shared
state.
.It Li i
Count operations for lines in the
.Dq Invalid
state.
.El
.Pp
If no
.Dq Li unitmask
qualifier is specified, the default is to count events for caches
lines in any of the above states.
.It Li usr
Configure the PMC to count events occurring at privilege levels 1, 2
or 3.
.El
.Pp
If neither of the
.Dq Li os
or
.Dq Li usr
qualifiers were specified, the default is to enable both.
.Pp
The event specifiers supported on AMD K7 PMCs are:
.Bl -tag -width indent
.It Li k7-dc-accesses
Count data cache accesses.
.It Li k7-dc-misses
Count data cache misses.
.It Li k7-dc-refills-from-l2 Op Li ,unitmask= Ns Ar mask
Count data cache refills from L2 cache.
This event may be further qualified using the
.Dq Li unitmask
qualifier.
.It Li k7-dc-refills-from-system Op Li ,unitmask= Ns Ar mask
Count data cache refills from system memory.
This event may be further qualified using the
.Dq Li unitmask
qualifier.
.It Li k7-dc-writebacks Op Li ,unitmask= Ns Ar mask
Count data cache writebacks.
This event may be further qualified using the
.Dq Li unitmask
qualifier.
.It Li k7-l1-dtlb-miss-and-l2-dtlb-hits
Count L1 DTLB misses and L2 DTLB hits.
.It Li k7-l1-and-l2-dtlb-misses
Count L1 and L2 DTLB misses.
.It Li k7-misaligned-references
Count misaligned data references.
.It Li k7-ic-fetches
Count instruction cache fetches.
.It Li k7-ic-misses
Count instruction cache misses.
.It Li k7-l1-itlb-misses
Count L1 ITLB misses that are L2 ITLB hits.
.It Li k7-l1-l2-itlb-misses
Count L1 (and L2) ITLB misses.
.It Li k7-retired-instructions
Count all retired instructions.
.It Li k7-retired-ops
Count retired ops.
.It Li k7-retired-branches
Count all retired branches (conditional, unconditional, exceptions
and interrupts).
.It Li k7-retired-branches-mispredicted
Count all misprediced retired branches.
.It Li k7-retired-taken-branches
Count retired taken branches.
.It Li k7-retired-taken-branches-mispredicted
Count mispredicted taken branches that were retired.
.It Li k7-retired-far-control-transfers
Count retired far control transfers.
.It Li k7-retired-resync-branches
Count retired resync branches (non control transfer branches).
.It Li k7-interrupts-masked-cycles
Count the number of cycles when the processor's
.Va IF
flag was zero.
.It Li k7-interrupts-masked-while-pending-cycles
Count the number of cycles interrupts were masked while pending due
to the processor's
.Va IF
flag being zero.
.It Li k7-hardware-interrupts
Count the number of taken hardware interrupts.
.El
.Ss AMD (K8) PMCs
These PMCs are present in the
.Tn "AMD Athlon64"
and
.Tn "AMD Opteron"
series of CPUs.
They are documented in:
.Rs
.%B "BIOS and Kernel Developer's Guide for the AMD Athlon(tm) 64 and AMD Opteron Processors"
.%N "Publication No. 26094"
.%D "April 2004"
.%Q "Advanced Micro Devices, Inc."
.Re
.Pp
Event specifiers for AMD K8 PMCs can have the following optional
qualifiers:
.Bl -tag -width indent
.It Li count= Ns Ar value
Configure the counter to increment only if the number of configured
events measured in a cycle is greater than or equal to
.Ar value .
.It Li edge
Configure the counter to only count negated-to-asserted transitions
of the conditions expressed by the other fields.
In other words, the counter will increment only once whenever a given
condition becomes true, irrespective of the number of clocks during
which the condition remains true.
.It Li inv
Invert the sense of comparision when the
.Dq Li count
qualifier is present, making the counter to increment when the
number of events per cycle is less than the value specified by
the
.Dq Li count
qualifier.
.It Li mask= Ns Ar qualifier
Many event specifiers for AMD K8 PMCs need to be additionally
qualified using a mask qualifier.
These additional qualifiers are event-specific and are documented
along with their associated event specifiers below.
.It Li os
Configure the PMC to count events happening at privilege level 0.
.It Li usr
Configure the PMC to count events occurring at privilege levels 1, 2
or 3.
.El
.Pp
If neither of the
.Dq Li os
or
.Dq Li usr
qualifiers were specified, the default is to enable both.
.Pp
The event specifiers supported on AMD K8 PMCs are:
.Bl -tag -width indent
.It Li k8-bu-cpu-clk-unhalted
Count the number of clock cycles when the CPU is not in the HLT or
STPCLK states.
.It Li k8-bu-fill-request-l2-miss Op Li ,mask= Ns Ar qualifier
Count fill requests that missed in the L2 cache.
This event may be further qualified using
.Ar qualifier ,
which is a
.Ql +
separated set of the following keywords:
.Pp
.Bl -tag -width indent -compact
.It Li dc-fill
Count data cache fill requests.
.It Li ic-fill
Count instruction cache fill requests.
.It Li tlb-reload
Count TLB reloads.
.El
.Pp
The default is to count all types of requests.
.It Li k8-bu-internal-l2-request Op Li ,mask= Ns Ar qualifier
Count internally generated requests to the L2 cache.
This event may be further qualified using
.Ar qualifier ,
which is a
.Ql +
separated set of the following keywords:
.Pp
.Bl -tag -width indent -compact
.It Li cancelled
Count cancelled requests.
.It Li dc-fill
Count data cache fill requests.
.It Li ic-fill
Count instruction cache fill requests.
.It Li tag-snoop
Count tag snoop requests.
.It Li tlb-reload
Count TLB reloads.
.El
.Pp
The default is to count all types of requests.
.It Li k8-dc-access
Count data cache accesses including microcode scratchpad accesses.
.It Li k8-dc-copyback Op Li ,mask= Ns Ar qualifier
Count data cache copyback operations.
This event may be further qualified using
.Ar qualifier ,
which is a
.Ql +
separated set of the following keywords:
.Pp
.Bl -tag -width indent -compact
.It Li exclusive
Count operations for lines in the
.Dq exclusive
state.
.It Li invalid
Count operations for lines in the
.Dq invalid
state.
.It Li modified
Count operations for lines in the
.Dq modified
state.
.It Li owner
Count operations for lines in the
.Dq owner
state.
.It Li shared
Count operations for lines in the
.Dq shared
state.
.El
.Pp
The default is to count operations for lines in all the
above states.
.It Li k8-dc-dcache-accesses-by-locks Op Li ,mask= Ns Ar qualifier
Count data cache accesses by lock instructions.
This event is only available on processors of revision C or later
vintage.
This event may be further qualified using
.Ar qualifier ,
which is a
.Ql +
separated set of the following keywords:
.Pp
.Bl -tag -width indent -compact
.It Li accesses
Count data cache accesses by lock instructions.
.It Li misses
Count data cache misses by lock instructions.
.El
.Pp
The default is to count all accesses.
.It Li k8-dc-dispatched-prefetch-instructions Op Li ,mask= Ns Ar qualifier
Count the number of dispatched prefetch instructions.
This event may be further qualified using
.Ar qualifier ,
which is a
.Ql +
separated set of the following keywords:
.Pp
.Bl -tag -width indent -compact
.It Li load
Count load operations.
.It Li nta
Count non-temporal operations.
.It Li store
Count store operations.
.El
.Pp
The default is to count all operations.
.It Li k8-dc-l1-dtlb-miss-and-l2-dtlb-hit
Count L1 DTLB misses that are L2 DTLB hits.
.It Li k8-dc-l1-dtlb-miss-and-l2-dtlb-miss
Count L1 DTLB misses that are also misses in the L2 DTLB.
.It Li k8-dc-microarchitectural-early-cancel-of-an-access
Count microarchitectural early cancels of data cache accesses.
.It Li k8-dc-microarchitectural-late-cancel-of-an-access
Count microarchitectural late cancels of data cache accesses.
.It Li k8-dc-misaligned-data-reference
Count misaligned data references.
.It Li k8-dc-miss
Count data cache misses.
.It Li k8-dc-one-bit-ecc-error Op Li ,mask= Ns Ar qualifier
Count one bit ECC errors found by the scrubber.
This event may be further qualified using
.Ar qualifier ,
which is a
.Ql +
separated set of the following keywords:
.Pp
.Bl -tag -width indent -compact
.It Li scrubber
Count scrubber detected errors.
.It Li piggyback
Count piggyback scrubber errors.
.El
.Pp
The default is to count both kinds of errors.
.It Li k8-dc-refill-from-l2 Op Li ,mask= Ns Ar qualifier
Count data cache refills from L2 cache.
This event may be further qualified using
.Ar qualifier ,
which is a
.Ql +
separated set of the following keywords:
.Pp
.Bl -tag -width indent -compact
.It Li exclusive
Count operations for lines in the
.Dq exclusive
state.
.It Li invalid
Count operations for lines in the
.Dq invalid
state.
.It Li modified
Count operations for lines in the
.Dq modified
state.
.It Li owner
Count operations for lines in the
.Dq owner
state.
.It Li shared
Count operations for lines in the
.Dq shared
state.
.El
.Pp
The default is to count operations for lines in all the
above states.
.It Li k8-dc-refill-from-system Op Li ,mask= Ns Ar qualifier
Count data cache refills from system memory.
This event may be further qualified using
.Ar qualifier ,
which is a
.Ql +
separated set of the following keywords:
.Pp
.Bl -tag -width indent -compact
.It Li exclusive
Count operations for lines in the
.Dq exclusive
state.
.It Li invalid
Count operations for lines in the
.Dq invalid
state.
.It Li modified
Count operations for lines in the
.Dq modified
state.
.It Li owner
Count operations for lines in the
.Dq owner
state.
.It Li shared
Count operations for lines in the
.Dq shared
state.
.El
.Pp
The default is to count operations for lines in all the
above states.
.It Li k8-fp-dispatched-fpu-ops Op Li ,mask= Ns Ar qualifier
Count the number of dispatched FPU ops.
This event is supported in revision B and later CPUs.
This event may be further qualified using
.Ar qualifier ,
which is a
.Ql +
separated set of the following keywords:
.Pp
.Bl -tag -width indent -compact
.It Li add-pipe-excluding-junk-ops
Count add pipe ops excluding junk ops.
.It Li add-pipe-junk-ops
Count junk ops in the add pipe.
.It Li multiply-pipe-excluding-junk-ops
Count multiply pipe ops excluding junk ops.
.It Li multiply-pipe-junk-ops
Count junk ops in the multiply pipe.
.It Li store-pipe-excluding-junk-ops
Count store pipe ops excluding junk ops
.It Li store-pipe-junk-ops
Count junk ops in the store pipe.
.El
.Pp
The default is to count all types of ops.
.It Li k8-fp-cycles-with-no-fpu-ops-retired
Count cycles when no FPU ops were retired.
This event is supported in revision B and later CPUs.
.It Li k8-fp-dispatched-fpu-fast-flag-ops
Count dispatched FPU ops that use the fast flag interface.
This event is supported in revision B and later CPUs.
.It Li k8-fr-decoder-empty
Count cycles when there was nothing to dispatch (i.e., the decoder
was empty).
.It Li k8-fr-dispatch-stalls
Count all dispatch stalls.
.It Li k8-fr-dispatch-stall-for-segment-load
Count dispatch stalls for segment loads.
.It Li k8-fr-dispatch-stall-for-serialization
Count dispatch stalls for serialization.
.It Li k8-fr-dispatch-stall-from-branch-abort-to-retire
Count dispatch stalls from branch abort to retiral.
.It Li k8-fr-dispatch-stall-when-fpu-is-full
Count dispatch stalls when the FPU is full.
.It Li k8-fr-dispatch-stall-when-ls-is-full
Count dispatch stalls when the load/store unit is full.
.It Li k8-fr-dispatch-stall-when-reorder-buffer-is-full
Count dispatch stalls when the reorder buffer is full.
.It Li k8-fr-dispatch-stall-when-reservation-stations-are-full
Count dispatch stalls when reservation stations are full.
.It Li k8-fr-dispatch-stall-when-waiting-for-all-to-be-quiet
Count dispatch stalls when waiting for all to be quiet.
.\" XXX What does "waiting for all to be quiet" mean?
.It Li k8-fr-dispatch-stall-when-waiting-far-xfer-or-resync-branch-pending
Count dispatch stalls when a far control transfer or a resync branch
is pending.
.It Li k8-fr-fpu-exceptions Op Li ,mask= Ns Ar qualifier
Count FPU exceptions.
This event is supported in revision B and later CPUs.
This event may be further qualified using
.Ar qualifier ,
which is a
.Ql +
separated set of the following keywords:
.Pp
.Bl -tag -width indent -compact
.It Li sse-and-x87-microtraps
Count SSE and x87 microtraps.
.It Li sse-reclass-microfaults
Count SSE reclass microfaults
.It Li sse-retype-microfaults
Count SSE retype microfaults
.It Li x87-reclass-microfaults
Count x87 reclass microfaults.
.El
.Pp
The default is to count all types of exceptions.
.It Li k8-fr-interrupts-masked-cycles
Count cycles when interrupts were masked (by CPU RFLAGS field IF was zero).
.It Li k8-fr-interrupts-masked-while-pending-cycles
Count cycles while interrupts were masked while pending (i.e., cycles
when INTR was asserted while CPU RFLAGS field IF was zero).
.It Li k8-fr-number-of-breakpoints-for-dr0
Count the number of breakpoints for DR0.
.It Li k8-fr-number-of-breakpoints-for-dr1
Count the number of breakpoints for DR1.
.It Li k8-fr-number-of-breakpoints-for-dr2
Count the number of breakpoints for DR2.
.It Li k8-fr-number-of-breakpoints-for-dr3
Count the number of breakpoints for DR3.
.It Li k8-fr-retired-branches
Count retired branches including exceptions and interrupts.
.It Li k8-fr-retired-branches-mispredicted
Count mispredicted retired branches.
.It Li k8-fr-retired-far-control-transfers
Count retired far control transfers (which are always mispredicted).
.It Li k8-fr-retired-fastpath-double-op-instructions Op Li ,mask= Ns Ar qualifier
Count retired fastpath double op instructions.
This event is supported in revision B and later CPUs.
This event may be further qualified using
.Ar qualifier ,
which is a
.Ql +
separated set of the following keywords:
.Pp
.Bl -tag -width indent -compact
.It Li low-op-pos-0
Count instructions with the low op in position 0.
.It Li low-op-pos-1
Count instructions with the low op in position 1.
.It Li low-op-pos-2
Count instructions with the low op in position 2.
.El
.Pp
The default is to count all types of instructions.
.It Li k8-fr-retired-fpu-instructions Op Li ,mask= Ns Ar qualifier
Count retired FPU instructions.
This event is supported in revision B and later CPUs.
This event may be further qualified using
.Ar qualifier ,
which is a
.Ql +
separated set of the following keywords:
.Pp
.Bl -tag -width indent -compact
.It Li mmx-3dnow
Count MMX and 3DNow!\& instructions.
.It Li packed-sse-sse2
Count packed SSE and SSE2 instructions.
.It Li scalar-sse-sse2
Count scalar SSE and SSE2 instructions
.It Li x87
Count x87 instructions.
.El
.Pp
The default is to count all types of instructions.
.It Li k8-fr-retired-near-returns
Count retired near returns.
.It Li k8-fr-retired-near-returns-mispredicted
Count mispredicted near returns.
.It Li k8-fr-retired-resyncs
Count retired resyncs (non-control transfer branches).
.It Li k8-fr-retired-taken-hardware-interrupts
Count retired taken hardware interrupts.
.It Li k8-fr-retired-taken-branches
Count retired taken branches.
.It Li k8-fr-retired-taken-branches-mispredicted
Count retired taken branches that were mispredicted.
.It Li k8-fr-retired-taken-branches-mispredicted-by-addr-miscompare
Count retired taken branches that were mispredicted only due to an
address miscompare.
.It Li k8-fr-retired-uops
Count retired uops.
.It Li k8-fr-retired-x86-instructions
Count retired x86 instructions including exceptions and interrupts.
.It Li k8-ic-fetch
Count instruction cache fetches.
.It Li k8-ic-instruction-fetch-stall
Count cycles in stalls due to instruction fetch.
.It Li k8-ic-l1-itlb-miss-and-l2-itlb-hit
Count L1 ITLB misses that are L2 ITLB hits.
.It Li k8-ic-l1-itlb-miss-and-l2-itlb-miss
Count ITLB misses that miss in both L1 and L2 ITLBs.
.It Li k8-ic-microarchitectural-resync-by-snoop
Count microarchitectural resyncs caused by snoops.
.It Li k8-ic-miss
Count instruction cache misses.
.It Li k8-ic-refill-from-l2
Count instruction cache refills from L2 cache.
.It Li k8-ic-refill-from-system
Count instruction cache refills from system memory.
.It Li k8-ic-return-stack-hits
Count hits to the return stack.
.It Li k8-ic-return-stack-overflow
Count overflows of the return stack.
.It Li k8-ls-buffer2-full
Count load/store buffer2 full events.
.It Li k8-ls-locked-operation Op Li ,mask= Ns Ar qualifier
Count locked operations.
For revision C and later CPUs, the following qualifiers are supported:
.Pp
.Bl -tag -width indent -compact
.It Li cycles-in-request
Count the number of cycles in the lock request/grant stage.
.It Li cycles-to-complete
Count the number of cycles a lock takes to complete once it is
non-speculative and is the older load/store operation.
.It Li locked-instructions
Count the number of lock instructions executed.
.El
.Pp
The default is to count the number of lock instructions executed.
.It Li k8-ls-microarchitectural-late-cancel
Count microarchitectural late cancels of operations in the load/store
unit.
.It Li k8-ls-microarchitectural-resync-by-self-modifying-code
Count microarchitectural resyncs caused by self-modifying code.
.It Li k8-ls-microarchitectural-resync-by-snoop
Count microarchitectural resyncs caused by snoops.
.It Li k8-ls-retired-cflush-instructions
Count retired CFLUSH instructions.
.It Li k8-ls-retired-cpuid-instructions
Count retired CPUID instructions.
.It Li k8-ls-segment-register-load Op Li ,mask= Ns Ar qualifier
Count segment register loads.
This event may be further qualified using
.Ar qualifier ,
which is a
.Ql +
separated set of the following keywords:
.Bl -tag -width indent -compact
.It Li cs
Count CS register loads.
.It Li ds
Count DS register loads.
.It Li es
Count ES register loads.
.It Li fs
Count FS register loads.
.It Li gs
Count GS register loads.
.\" .It Li hs
.\" Count HS register loads.
.\" XXX "HS" register?
.It Li ss
Count SS register loads.
.El
.Pp
The default is to count all types of loads.
.It Li k8-nb-memory-controller-bypass-saturation Op Li ,mask= Ns Ar qualifier
Count memory controller bypass counter saturation events.
This event may be further qualified using
.Ar qualifier ,
which is a
.Ql +
separated set of the following keywords:
.Pp
.Bl -tag -width indent -compact
.It Li dram-controller-interface-bypass
Count DRAM controller interface bypass.
.It Li dram-controller-queue-bypass
Count DRAM controller queue bypass.
.It Li memory-controller-hi-pri-bypass
Count memory controller high priority bypasses.
.It Li memory-controller-lo-pri-bypass
Count memory controller low priority bypasses.
.El
.Pp
.It Li k8-nb-memory-controller-dram-slots-missed
Count memory controller DRAM command slots missed (in MemClks).
.It Li k8-nb-memory-controller-page-access-event Op Li ,mask= Ns Ar qualifier
Count memory controller page access events.
This event may be further qualified using
.Ar qualifier ,
which is a
.Ql +
separated set of the following keywords:
.Pp
.Bl -tag -width indent -compact
.It Li page-conflict
Count page conflicts.
.It Li page-hit
Count page hits.
.It Li page-miss
Count page misses.
.El
.Pp
The default is to count all types of events.
.It Li k8-nb-memory-controller-page-table-overflow
Count memory control page table overflow events.
.It Li k8-nb-probe-result Op Li ,mask= Ns Ar qualifier
Count probe events.
This event may be further qualified using
.Ar qualifier ,
which is a
.Ql +
separated set of the following keywords:
.Pp
.Bl -tag -width indent -compact
.It Li probe-hit
Count all probe hits.
.It Li probe-hit-dirty-no-memory-cancel
Count probe hits without memory cancels.
.It Li probe-hit-dirty-with-memory-cancel
Count probe hits with memory cancels.
.It Li probe-miss
Count probe misses.
.El
.It Li k8-nb-sized-commands Op Li ,mask= Ns Ar qualifier
Count sized commands issued.
This event may be further qualified using
.Ar qualifier ,
which is a
.Ql +
separated set of the following keywords:
.Pp
.Bl -tag -width indent -compact
.It Li nonpostwrszbyte
.It Li nonpostwrszdword
.It Li postwrszbyte
.It Li postwrszdword
.It Li rdszbyte
.It Li rdszdword
.It Li rdmodwr
.El
.Pp
The default is to count all types of commands.
.It Li k8-nb-memory-controller-turnaround Op Li ,mask= Ns Ar qualifier
Count memory control turnaround events.
This event may be further qualified using
.Ar qualifier ,
which is a
.Ql +
separated set of the following keywords:
.Pp
.Bl -tag -width indent -compact
.\" XXX doc is unclear whether these are cycle counts or event counts
.It Li dimm-turnaround
Count DIMM turnarounds.
.It Li read-to-write-turnaround
Count read to write turnarounds.
.It Li write-to-read-turnaround
Count write to read turnarounds.
.El
.Pp
The default is to count all types of events.
.It Li k8-nb-ht-bus0-bandwidth Op Li ,mask= Ns Ar qualifier
.It Li k8-nb-ht-bus1-bandwidth Op Li ,mask= Ns Ar qualifier
.It Li k8-nb-ht-bus2-bandwidth Op Li ,mask= Ns Ar qualifier
Count events on the HyperTransport(tm) buses.
These events may be further qualified using
.Ar qualifier ,
which is a
.Ql +
separated set of the following keywords:
.Pp
.Bl -tag -width indent -compact
.It Li buffer-release
Count buffer release messages sent.
.It Li command
Count command messages sent.
.It Li data
Count data messages sent.
.It Li nop
Count nop messages sent.
.El
.Pp
The default is to count all types of messages.
.El
.Ss Intel Pentium PMCS
Intel Pentium PMCs are present in Intel
.Tn Pentium
and
.Tn "Pentium MMX"
processors.
.Pp
These CPUs have two counters.
Some events may only be used on specific counters and some events
are defined only on processors supporting the MMX instruction set.
.Pp
These PMCs are documented in
.Rs
.%B "Intel 64 and IA-32 Intel(R) Architectures Software Developer's Manual"
.%T "Volume 3B: System Programming Guide, Part 2"
.%N "Order Number 253669-024US"
.%D "August 2007"
.%Q "Intel Corporation"
.Re
.Pp
Event specifiers for Intel Pentium PMCs can have the following common
qualifiers:
.Bl -tag -width indent
.It Li duration
Count duration (in clocks) of events.
The default is to count events.
.It Li os
Measure events at privilege levels 0, 1 and 2.
.It Li overflow
Assert the external processor pin associated with a counter on counter
overflow.
.It Li usr
Measure events at privilege level 3.
.El
.Pp
Note that these PMCs do not have the ability to interrupt the CPU.
.Pp
The event specifiers supported by Intel Pentium PMCs are:
.Bl -tag -width indent
.It Li p5-any-segment-register-loaded
The number of writes to any segment register, including the LDTR,
GDTR, TR and IDTR.
Far control transfers and task switches that involve privilege
level changes will count this event twice.
.It Li p5-bank-conflicts
The number of actual bank conflicts.
.It Li p5-branches
The number of taken and not taken branches including branches, jumps, calls,
software interrupts and interrupt returns.
.It Li p5-breakpoint-match-on-dr0-register
The number of matches on the DR0 breakpoint register.
.It Li p5-breakpoint-match-on-dr1-register
The number of matches on the DR1 breakpoint register.
.It Li p5-breakpoint-match-on-dr2-register
The number of matches on the DR2 breakpoint register.
.It Li p5-breakpoint-match-on-dr3-register
The number of matches on the DR3 breakpoint register.
.It Li p5-btb-false-entries
.Pq Tn Pentium MMX
The number of false entries in the BTB.
This event is only allocated on counter 0.
.It Li p5-btb-hits
The number of branches executed that hit in the branch table buffer.
.It Li p5-btb-miss-prediction-on-not-taken-branch
.Pq Tn Pentium MMX
The number of times the BTB predicted a not-taken branch as taken.
This event is only allocated on counter 1.
.It Li p5-bus-cycle-duration
The number of cycles while a bus cycle was in progress.
.It Li p5-bus-ownership-latency
.Pq Tn Pentium MMX
The time from bus ownership being requested to ownership being granted.
This event is only allocated on counter 0.
.It Li p5-bus-ownership-transfers
.Pq Tn Pentium MMX
The number of bus ownership transfers.
This event is only allocated on counter 1.
.It Li p5-bus-utilization-due-to-processor-activity
.Pq Tn Pentium MMX
The number of clocks the bus is busy due to the processor's own
activity.
This event is only allocated on counter 0.
.It Li p5-cache-line-sharing
.Pq Tn Pentium MMX
The number of shared data lines in L1 cache.
This event is only allocated on counter 1.
.It Li p5-cache-m-state-line-sharing
.Pq Tn Pentium MMX
The number of hits to an M- state line due to a memory access by
another processor.
This event is only allocated on counter 0.
.It Li p5-code-cache-miss
The number of instruction reads that miss the internal code cache.
Both cacheable and uncacheable misses are counted.
.It Li p5-code-read
The number of instruction reads to both cacheable and uncacheable regions.
.It Li p5-code-tlb-miss
The number of instruction reads that miss the instruction TLB.
Both cacheable and uncacheable unreads are counted.
.It Li p5-d1-starvation-and-fifo-is-empty
.Pq Tn Pentium MMX
The number of times the D1 stage cannot issue any instructions because
the FIFO was empty.
This event is only allocated on counter 0.
.It Li p5-d1-starvation-and-only-one-instruction-in-fifo
.Pq Tn Pentium MMX
The number of times the D1 stage could issue only one instruction
because the FIFO had one instruction ready.
This event is only allocated on counter 1.
.It Li p5-data-cache-lines-written-back
The number of data cache lines that are written back, including
those caused by internal and external snoops.
.It Li p5-data-cache-tlb-miss-stall-duration
.Pq Tn Pentium MMX
The number of clocks the pipeline is stalled due to a data cache
TLB miss.
This event is only allocated on counter 1.
.It Li p5-data-read
The number of memory data reads, counting internal data cache hits and
misses.
I/O and data memory accesses due to TLB miss processing are
not included.
Split cycle reads are counted individually.
.It Li p5-data-read-miss
The number of memory read accesses that miss the data cache, counting
both cacheable and uncacheable accesses.
Data accesses that are part of TLB miss processing are not included.
I/O accesses are not included.
.It Li p5-data-read-miss-or-write-miss
The number of data reads and writes that miss the internal data cache,
counting uncacheable accesses.
Data accesses due to TLB miss processing are not counted.
.It Li p5-data-read-or-write
The number of data reads and writes including internal data cache hits
and misses.
Data reads due to TLB miss processing are not counted.
.It Li p5-data-tlb-miss
The number of misses to the data cache translation lookaside buffer.
.It Li p5-data-write
The number of memory data writes, counting internal data cache hits
and misses.
I/O is not included and split cycle writes are counted individually.
.It Li p5-data-write-miss
The number of memory write accesses that miss the data cache, counting
both cacheable and uncacheable accesses.
I/O accesses are not counted.
.It Li p5-emms-instructions-executed
.Pq Tn Pentium MMX
The number of EMMS instructions executed.
This event is only allocated on counter 0.
.It Li p5-external-data-cache-snoop-hits
The number of external snoops to the data cache that hit a valid line,
or the data line fill buffer, or one of the write back buffers.
.It Li p5-external-snoops
The number of external snoop requests accepted, including snoops that
hit in the code cache, the data cache and that hit in neither.
.It Li p5-floating-point-stalls-duration
.Pq Tn Pentium MMX
The number of cycles the pipeline is stalled due to a floating point
freeze.
This event is only allocated on counter 0.
.It Li p5-flops
The number of floating point adds, subtracts, multiples, divides and
square roots.
Transcendental instructions trigger this event multiple times.
Instructions generating divide-by-zero, negative square root, special
operand and stack exceptions are not counted.
Integer multiply instructions that use the x87 FPU are counted.
.It Li p5-full-write-buffer-stall-duration-while-executing-mmx-instructions
.Pq Tn Pentium MMX
The number of clocks the pipeline has stalled due to full write
buffers when executing MMX instructions.
This event is only allocated on counter 0.
.It Li p5-hardware-interrupts
The number of taken INTR and NMI interrupts.
.It Li p5-instructions-executed
The number of instructions executed.
Repeat prefixed instructions are counted only once.
The HLT instruction is counted only once, irrespective of the number
of cycles spent in the halted state.
All hardware and software exceptions are counted as instructions, and
fault handler invocations are also counted as instructions.
.It Li p5-instructions-executed-v-pipe
The number of instructions that executed in the V pipe.
.It Li p5-io-read-or-write-cycle
The number of bus cycles directed to I/O space.
.It Li p5-locked-bus-cycle
The number of locked bus cycles that occur on account of the lock
prefixes, LOCK instructions, page table updates and descriptor table
updates.
.It Li p5-memory-accesses-in-both-pipes
The number of data memory reads or writes that are paired in both pipes.
.It Li p5-misaligned-data-memory-on-mmx-instructions
.Pq Tn Pentium MMX
The number of misaligned data memory references when executing MMX
instructions.
This event is only allocated on counter 0.
.It Li p5-misaligned-data-memory-or-io-references
The number of memory or I/O reads or writes that are not aligned on
natural boundaries.
2- and 4-byte accesses are counted as misaligned if they cross a 4
byte boundary.
.It Li p5-mispredicted-or-unpredicted-returns
.Pq Tn Pentium MMX
The number of returns predicted incorrectly or not at all, only
counting RET instructions.
This event is only allocated on counter 0.
.It Li p5-mmx-instruction-data-read-misses
.Pq Tn Pentium MMX
The number of MMX instruction data read misses.
This event is only allocated on counter 1.
.It Li p5-mmx-instruction-data-reads
.Pq Tn Pentium MMX
The number of MMX instruction data reads.
This event is only allocated on counter 0.
.It Li p5-mmx-instruction-data-write-misses
.Pq Tn Pentium MMX
The number of data write misses caused by MMX instructions.
This event is only allocated on counter 1.
.It Li p5-mmx-instruction-data-writes
.Pq Tn Pentium MMX
The number of data writes caused by MMX instructions.
This event is only allocated on counter 0.
.It Li p5-mmx-instructions-executed-u-pipe
.Pq Tn Pentium MMX
The number of MMX instructions executed in the U pipe.
This event is only allocated on counter 0.
.It Li p5-mmx-instructions-executed-v-pipe
The number of MMX instructions executed in the V pipe.
This event is only allocated on counter 1.
.It Li p5-mmx-multiply-unit-interlock
.Pq Tn Pentium MMX
The number of clocks the pipeline is stalled because the destination
of a prior MMX multiply is not ready.
This event is only allocated on counter 0.
.It Li p5-movd-movq-store-stall-due-to-previous-mmx-operation
.Pq Tn Pentium MMX
The number of clocks a MOVD/MOVQ instruction stalled in the D2 stage
of the pipeline due to a previous MMX instruction.
This event is only allocated on counter 1.
.It Li p5-noncacheable-memory-reads
The number of bus cycles for non-cacheable instruction or data reads,
including cycles caused by TLB misses.
.It Li p5-number-of-cycles-not-in-halt-state
.Pq Tn Pentium MMX
The number of cycles the processor is not idle due to the HLT
instruction.
This event is only allocated on counter 0.
.It Li p5-pipeline-agi-stalls
The number of address generation interlock stalls.
An AGI that occurs in both the U and V pipelines in the same clock
signals the event twice.
.It Li p5-pipeline-flushes
The number of pipeline flushes that occur.
Pipeline flushes are caused by branch mispredicts, exceptions,
interrupts, some segment register loads, and BTB misses.
Prefetch queue flushes due to serializing instructions are not
counted.
.It Li p5-pipeline-flushes-due-to-wrong-branch-predictions
.Pq Tn Pentium MMX
The number of pipeline flushes due to wrong branch predictions
resolved in either the E- or WB- stage of the pipeline.
This event is only allocated on counter 0.
.It Li p5-pipeline-flushes-due-to-wrong-branch-predictions-resolved-in-wb-stage
.Pq Tn Pentium MMX
The number of pipeline flushes due to wrong branch predictions
resolved in the stage of the pipeline.
This event is only allocated on counter 1.
.It Li p5-pipeline-stall-for-mmx-instruction-data-memory-reads
.Pq Tn Pentium MMX
The number of clocks during pipeline stalls caused by waiting MMX data
memory reads.
This event is only allocated on counter 0.
.It Li p5-predicted-returns
.Pq Tn Pentium MMX
The number of predicted returns, whether correct or incorrect.
This counter only counts RET instructions.
This event is only allocated on counter 1.
.It Li p5-returns
.Pq Tn Pentium MMX
The number of RET instructions executed.
This event is only allocated on counter 0.
.It Li p5-saturating-mmx-instructions-executed
.Pq Tn Pentium MMX
The number of saturating MMX instructions executed.
This event is only allocated on counter 0.
.It Li p5-saturations-performed
.Pq Tn Pentium MMX
The number of saturating MMX instructions executed when at least one
of its results were actually saturated.
This event is only allocated on counter 1.
.It Li p5-stall-on-mmx-instruction-write-to-e-o-m-state-line
.Pq Tn Pentium MMX
The number of clocks during stalls on MMX instructions writing to
E- or M- state cache lines.
This event is only allocated on counter 1.
.It Li p5-stall-on-write-to-an-e-or-m-state-line
The number of stalls on a write to an exclusive or modified data cache
line.
.It Li p5-taken-branch-or-btb-hit
The number of events that may cause a hit in the BTB, namely either
taken branches or BTB hits.
.It Li p5-taken-branches
.Pq Tn Pentium MMX
The number of taken branches.
This event is only allocated on counter 1.
.It Li p5-transitions-between-mmx-and-fp-instructions
.Pq Tn Pentium MMX
The number of transitions between MMX and floating-point instructions
and vice-versa.
This event is only allocated on counter 1.
.It Li p5-waiting-for-data-memory-read-stall-duration
The number of clocks the pipeline was stalled waiting for data
memory reads.
Data TLB misses processing is included in this count.
.It Li p5-write-buffer-full-stall-duration
The number of clocks while the pipeline was stalled due to write
buffers being full.
.It Li p5-write-hit-to-m-or-e-state-lines
The number of writes that hit exclusive or modified lines in the data
cache.
.It Li p5-writes-to-noncacheable-memory
.Pq Tn Pentium MMX
The number of writes to non-cacheable memory, including write cycles
caused by TLB misses and I/O writes.
This event is only allocated on counter 1.
.El
.Ss Intel P6 PMCS
Intel P6 PMCs are present in Intel
.Tn "Pentium Pro" ,
.Tn "Pentium II" ,
.Tn Celeron ,
.Tn "Pentium III"
and
.Tn "Pentium M"
processors.
.Pp
These CPUs have two counters.
Some events may only be used on specific counters and some events are
defined only on specific processor models.
.Pp
These PMCs are documented in
.Rs
.%B "IA-32 Intel(R) Architecture Software Developer's Manual"
.%T "Volume 3: System Programming Guide"
.%N "Order Number 245472-012"
.%D 2003
.%Q "Intel Corporation"
.Re
.Pp
Some of these events are affected by processor errata described in
.Rs
.%B "Intel(R) Pentium(R) III Processor Specification Update"
.%N "Document Number: 244453-054"
.%D "April 2005"
.%Q "Intel Corporation"
.Re
.Pp
Event specifiers for Intel P6 PMCs can have the following common
qualifiers:
.Bl -tag -width indent
.It Li cmask= Ns Ar value
Configure the PMC to increment only if the number of configured
events measured in a cycle is greater than or equal to
.Ar value .
.It Li edge
Configure the PMC to count the number of deasserted to asserted
transitions of the conditions expressed by the other qualifiers.
If specified, the counter will increment only once whenever a
condition becomes true, irrespective of the number of clocks during
which the condition remains true.
.It Li inv
Invert the sense of comparision when the
.Dq Li cmask
qualifier is present, making the counter increment when the number of
events per cycle is less than the value specified by the
.Dq Li cmask
qualifier.
.It Li os
Configure the PMC to count events happening at processor privilege
level 0.
.It Li umask= Ns Ar value
This qualifier is used to further qualify the event selected (see
below).
.It Li usr
Configure the PMC to count events occurring at privilege levels 1, 2
or 3.
.El
.Pp
If neither of the
.Dq Li os
or
.Dq Li usr
qualifiers are specified, the default is to enable both.
.Pp
The event specifiers supported by Intel P6 PMCs are:
.Bl -tag -width indent
.It Li p6-baclears
Count the number of times a static branch prediction was made by the
branch decoder because the BTB did not have a prediction.
.It Li p6-br-bac-missp-exec
.Pq Tn "Pentium M"
Count the number of branch instructions executed that where
mispredicted at the Front End (BAC).
.It Li p6-br-bogus
Count the number of bogus branches.
.It Li p6-br-call-exec
.Pq Tn "Pentium M"
Count the number of call instructions executed.
.It Li p6-br-call-missp-exec
.Pq Tn "Pentium M"
Count the number of call instructions executed that were mispredicted.
.It Li p6-br-cnd-exec
.Pq Tn "Pentium M"
Count the number of conditional branch instructions executed.
.It Li p6-br-cnd-missp-exec
.Pq Tn "Pentium M"
Count the number of conditional branch instructions executed that were
mispredicted.
.It Li p6-br-ind-call-exec
.Pq Tn "Pentium M"
Count the number of indirect call instructions executed.
.It Li p6-br-ind-exec
.Pq Tn "Pentium M"
Count the number of indirect branch instructions executed.
.It Li p6-br-ind-missp-exec
.Pq Tn "Pentium M"
Count the number of indirect branch instructions executed that were
mispredicted.
.It Li p6-br-inst-decoded
Count the number of branch instructions decoded.
.It Li p6-br-inst-exec
.Pq Tn "Pentium M"
Count the number of branch instructions executed but necessarily retired.
.It Li p6-br-inst-retired
Count the number of branch instructions retired.
.It Li p6-br-miss-pred-retired
Count the number of mispredicted branch instructions retired.
.It Li p6-br-miss-pred-taken-ret
Count the number of taken mispredicted branches retired.
.It Li p6-br-missp-exec
.Pq Tn "Pentium M"
Count the number of branch instructions executed that were
mispredicted at execution.
.It Li p6-br-ret-bac-missp-exec
.Pq Tn "Pentium M"
Count the number of return instructions executed that were
mispredicted at the Front End (BAC).
.It Li p6-br-ret-exec
.Pq Tn "Pentium M"
Count the number of return instructions executed.
.It Li p6-br-ret-missp-exec
.Pq Tn "Pentium M"
Count the number of return instructions executed that were
mispredicted at execution.
.It Li p6-br-taken-retired
Count the number of taken branches retired.
.It Li p6-btb-misses
Count the number of branches for which the BTB did not produce a
prediction.
.It Li p6-bus-bnr-drv
Count the number of bus clock cycles during which this processor is
driving the BNR# pin.
.It Li p6-bus-data-rcv
Count the number of bus clock cycles during which this processor is
receiving data.
.It Li p6-bus-drdy-clocks Op Li ,umask= Ns Ar qualifier
Count the number of clocks during which DRDY# is asserted.
An additional qualifier may be specified, and comprises one of the
following keywords:
.Pp
.Bl -tag -width indent -compact
.It Li any
Count transactions generated by any agent on the bus.
.It Li self
Count transactions generated by this processor.
.El
.Pp
The default is to count operations generated by this processor.
.It Li p6-bus-hit-drv
Count the number of bus clock cycles during which this processor is
driving the HIT# pin.
.It Li p6-bus-hitm-drv
Count the number of bus clock cycles during which this processor is
driving the HITM# pin.
.It Li p6-bus-lock-clocks Op Li ,umask= Ns Ar qualifier
Count the number of clocks during with LOCK# is asserted on the
external system bus.
An additional qualifier may be specified and comprises one of the following
keywords:
.Pp
.Bl -tag -width indent -compact
.It Li any
Count transactions generated by any agent on the bus.
.It Li self
Count transactions generated by this processor.
.El
.Pp
The default is to count operations generated by this processor.
.It Li p6-bus-req-outstanding
Count the number of bus requests outstanding in any given cycle.
.It Li p6-bus-snoop-stall
Count the number of clock cycles during which the bus is snoop stalled.
.It Li p6-bus-tran-any Op Li ,umask= Ns Ar qualifier
Count the number of completed bus transactions of any kind.
An additional qualifier may be specified and comprises one of the following
keywords:
.Pp
.Bl -tag -width indent -compact
.It Li any
Count transactions generated by any agent on the bus.
.It Li self
Count transactions generated by this processor.
.El
.Pp
The default is to count operations generated by this processor.
.It Li p6-bus-tran-brd Op Li ,umask= Ns Ar qualifier
Count the number of burst read transactions.
An additional qualifier may be specified and comprises one of the following
keywords:
.Pp
.Bl -tag -width indent -compact
.It Li any
Count transactions generated by any agent on the bus.
.It Li self
Count transactions generated by this processor.
.El
.Pp
The default is to count operations generated by this processor.
.It Li p6-bus-tran-burst Op Li ,umask= Ns Ar qualifier
Count the number of completed burst transactions.
An additional qualifier may be specified and comprises one of the following
keywords:
.Pp
.Bl -tag -width indent -compact
.It Li any
Count transactions generated by any agent on the bus.
.It Li self
Count transactions generated by this processor.
.El
.Pp
The default is to count operations generated by this processor.
.It Li p6-bus-tran-def Op Li ,umask= Ns Ar qualifier
Count the number of completed deferred transactions.
An additional qualifier may be specified and comprises one of the following
keywords:
.Pp
.Bl -tag -width indent -compact
.It Li any
Count transactions generated by any agent on the bus.
.It Li self
Count transactions generated by this processor.
.El
.Pp
The default is to count operations generated by this processor.
.It Li p6-bus-tran-ifetch Op Li ,umask= Ns Ar qualifier
Count the number of completed instruction fetch transactions.
An additional qualifier may be specified and comprises one of the following
keywords:
.Pp
.Bl -tag -width indent -compact
.It Li any
Count transactions generated by any agent on the bus.
.It Li self
Count transactions generated by this processor.
.El
.Pp
The default is to count operations generated by this processor.
.It Li p6-bus-tran-inval Op Li ,umask= Ns Ar qualifier
Count the number of completed invalidate transactions.
An additional qualifier may be specified and comprises one of the following
keywords:
.Pp
.Bl -tag -width indent -compact
.It Li any
Count transactions generated by any agent on the bus.
.It Li self
Count transactions generated by this processor.
.El
.Pp
The default is to count operations generated by this processor.
.It Li p6-bus-tran-mem Op Li ,umask= Ns Ar qualifier
Count the number of completed memory transactions.
An additional qualifier may be specified and comprises one of the following
keywords:
.Pp
.Bl -tag -width indent -compact
.It Li any
Count transactions generated by any agent on the bus.
.It Li self
Count transactions generated by this processor.
.El
.Pp
The default is to count operations generated by this processor.
.It Li p6-bus-tran-pwr Op Li ,umask= Ns Ar qualifier
Count the number of completed partial write transactions.
An additional qualifier may be specified and comprises one of the following
keywords:
.Pp
.Bl -tag -width indent -compact
.It Li any
Count transactions generated by any agent on the bus.
.It Li self
Count transactions generated by this processor.
.El
.Pp
The default is to count operations generated by this processor.
.It Li p6-bus-tran-rfo Op Li ,umask= Ns Ar qualifier
Count the number of completed read-for-ownership transactions.
An additional qualifier may be specified and comprises one of the following
keywords:
.Pp
.Bl -tag -width indent -compact
.It Li any
Count transactions generated by any agent on the bus.
.It Li self
Count transactions generated by this processor.
.El
.Pp
The default is to count operations generated by this processor.
.It Li p6-bus-trans-io Op Li ,umask= Ns Ar qualifier
Count the number of completed I/O transactions.
An additional qualifier may be specified and comprises one of the following
keywords:
.Pp
.Bl -tag -width indent -compact
.It Li any
Count transactions generated by any agent on the bus.
.It Li self
Count transactions generated by this processor.
.El
.Pp
The default is to count operations generated by this processor.
.It Li p6-bus-trans-p Op Li ,umask= Ns Ar qualifier
Count the number of completed partial transactions.
An additional qualifier may be specified and comprises one of the following
keywords:
.Pp
.Bl -tag -width indent -compact
.It Li any
Count transactions generated by any agent on the bus.
.It Li self
Count transactions generated by this processor.
.El
.Pp
The default is to count operations generated by this processor.
.It Li p6-bus-trans-wb Op Li ,umask= Ns Ar qualifier
Count the number of completed write-back transactions.
An additional qualifier may be specified and comprises one of the following
keywords:
.Pp
.Bl -tag -width indent -compact
.It Li any
Count transactions generated by any agent on the bus.
.It Li self
Count transactions generated by this processor.
.El
.Pp
The default is to count operations generated by this processor.
.It Li p6-cpu-clk-unhalted
Count the number of cycles during with the processor was not halted.
.Pp
.Pq Tn "Pentium M"
Count the number of cycles during with the processor was not halted
and not in a thermal trip.
.It Li p6-cycles-div-busy
Count the number of cycles during which the divider is busy and cannot
accept new divides.
This event is only allocated on counter 0.
.It Li p6-cycles-in-pending-and-masked
Count the number of processor cycles for which interrupts were
disabled and interrupts were pending.
.It Li p6-cycles-int-masked
Count the number of processor cycles for which interrupts were
disabled.
.It Li p6-data-mem-refs
Count all loads and all stores using any memory type, including
internal retries.
Each part of a split store is counted separately.
.It Li p6-dcu-lines-in
Count the total lines allocated in the data cache unit.
.It Li p6-dcu-m-lines-in
Count the number of M state lines allocated in the data cache unit.
.It Li p6-dcu-m-lines-out
Count the number of M state lines evicted from the data cache unit.
.It Li p6-dcu-miss-outstanding
Count the weighted number of cycles while a data cache unit miss is
outstanding, incremented by the number of outstanding cache misses at
any time.
.It Li p6-div
Count the number of integer and floating-point divides including
speculative divides.
This event is only allocated on counter 1.
.It Li p6-emon-esp-uops
.Pq Tn "Pentium M"
Count the total number of micro-ops.
.It Li p6-emon-est-trans Op Li ,umask= Ns Ar qualifier
.Pq Tn "Pentium M"
Count the number of
.Tn "Enhanced Intel SpeedStep"
transitions.
An additional qualifier may be specified, and can be one of the
following keywords:
.Pp
.Bl -tag -width indent -compact
.It Li all
Count all transitions.
.It Li freq
Count only frequency transitions.
.El
.Pp
The default is to count all transitions.
.It Li p6-emon-fused-uops-ret Op Li ,umask= Ns Ar qualifier
.Pq Tn "Pentium M"
Count the number of retired fused micro-ops.
An additional qualifier may be specified, and may be one of the
following keywords:
.Pp
.Bl -tag -width indent -compact
.It Li all
Count all fused micro-ops.
.It Li loadop
Count only load and op micro-ops.
.It Li stdsta
Count only STD/STA micro-ops.
.El
.Pp
The default is to count all fused micro-ops.
.It Li p6-emon-kni-comp-inst-ret
.Pq Tn "Pentium III"
Count the number of SSE computational instructions retired.
An additional qualifier may be specified, and comprises one of the
following keywords:
.Pp
.Bl -tag -width indent -compact
.It Li packed-and-scalar
Count packed and scalar operations.
.It Li scalar
Count scalar operations only.
.El
.Pp
The default is to count packed and scalar operations.
.It Li p6-emon-kni-inst-retired Op Li ,umask= Ns Ar qualifier
.Pq Tn "Pentium III"
Count the number of SSE instructions retired.
An additional qualifier may be specified, and comprises one of the
following keywords:
.Pp
.Bl -tag -width indent -compact
.It Li packed-and-scalar
Count packed and scalar operations.
.It Li scalar
Count scalar operations only.
.El
.Pp
The default is to count packed and scalar operations.
.It Li p6-emon-kni-pref-dispatched Op Li ,umask= Ns Ar qualifier
.Pq Tn "Pentium III"
Count the number of SSE prefetch or weakly ordered instructions
dispatched (including speculative prefetches).
An additional qualifier may be specified, and comprises one of the
following keywords:
.Pp
.Bl -tag -width indent -compact
.It Li nta
Count non-temporal prefetches.
.It Li t1
Count prefetches to L1.
.It Li t2
Count prefetches to L2.
.It Li wos
Count weakly ordered stores.
.El
.Pp
The default is to count non-temporal prefetches.
.It Li p6-emon-kni-pref-miss Op Li ,umask= Ns Ar qualifier
.Pq Tn "Pentium III"
Count the number of prefetch or weakly ordered instructions that miss
all caches.
An additional qualifier may be specified, and comprises one of the
following keywords:
.Pp
.Bl -tag -width indent -compact
.It Li nta
Count non-temporal prefetches.
.It Li t1
Count prefetches to L1.
.It Li t2
Count prefetches to L2.
.It Li wos
Count weakly ordered stores.
.El
.Pp
The default is to count non-temporal prefetches.
.It Li p6-emon-pref-rqsts-dn
.Pq Tn "Pentium M"
Count the number of downward prefetches issued.
.It Li p6-emon-pref-rqsts-up
.Pq Tn "Pentium M"
Count the number of upward prefetches issued.
.It Li p6-emon-simd-instr-retired
.Pq Tn "Pentium M"
Count the number of retired
.Tn MMX
instructions.
.It Li p6-emon-sse-sse2-comp-inst-retired Op Li ,umask= Ns Ar qualifier
.Pq Tn "Pentium M"
Count the number of computational SSE instructions retired.
An additional qualifier may be specified and can be one of the
following keywords:
.Pp
.Bl -tag -width indent -compact
.It Li sse-packed-single
Count SSE packed-single instructions.
.It Li sse-scalar-single
Count SSE scalar-single instructions.
.It Li sse2-packed-double
Count SSE2 packed-double instructions.
.It Li sse2-scalar-double
Count SSE2 scalar-double instructions.
.El
.Pp
The default is to count SSE packed-single instructions.
.It Li p6-emon-sse-sse2-inst-retired Op Li ,umask= Ns Ar qualifer
.Pp
.Pq Tn "Pentium M"
Count the number of SSE instructions retired.
An additional qualifier can be specified, and can be one of the
following keywords:
.Pp
.Bl -tag -width indent -compact
.It Li sse-packed-single
Count SSE packed-single instructions.
.It Li sse-packed-single-scalar-single
Count SSE packed-single and scalar-single instructions.
.It Li sse2-packed-double
Count SSE2 packed-double instructions.
.It Li sse2-scalar-double
Count SSE2 scalar-double instructions.
.El
.Pp
The default is to count SSE packed-single instructions.
.It Li p6-emon-synch-uops
.Pq Tn "Pentium M"
Count the number of sync micro-ops.
.It Li p6-emon-thermal-trip
.Pq Tn "Pentium M"
Count the duration or occurrences of thermal trips.
Use the
.Dq Li edge
qualifier to count occurrences of thermal trips.
.It Li p6-emon-unfusion
.Pq Tn "Pentium M"
Count the number of unfusion events in the reorder buffer.
.It Li p6-flops
Count the number of computational floating point operations retired.
This event is only allocated on counter 0.
.It Li p6-fp-assist
Count the number of floating point exceptions handled by microcode.
This event is only allocated on counter 1.
.It Li p6-fp-comps-ops-exe
Count the number of computation floating point operations executed.
This event is only allocated on counter 0.
.It Li p6-fp-mmx-trans Op Li ,umask= Ns Ar qualifier
.Pq Tn "Pentium II" , Tn "Pentium III"
Count the number of transitions between MMX and floating-point
instructions.
An additional qualifier may be specified, and comprises one of the
following keywords:
.Pp
.Bl -tag -width indent -compact
.It Li mmxtofp
Count transitions from MMX instructions to floating-point instructions.
.It Li fptommx
Count transitions from floating-point instructions to MMX instructions.
.El
.Pp
The default is to count MMX to floating-point transitions.
.It Li p6-hw-int-rx
Count the number of hardware interrupts received.
.It Li p6-ifu-fetch
Count the number of instruction fetches, both cacheable and non-cacheable.
.It Li p6-ifu-fetch-miss
Count the number of instruction fetch misses (i.e., those that produce
memory accesses).
.It Li p6-ifu-mem-stall
Count the number of cycles instruction fetch is stalled for any reason.
.It Li p6-ild-stall
Count the number of cycles the instruction length decoder is stalled.
.It Li p6-inst-decoded
Count the number of instructions decoded.
.It Li p6-inst-retired
Count the number of instructions retired.
.It Li p6-itlb-miss
Count the number of instruction TLB misses.
.It Li p6-l2-ads
Count the number of L2 address strobes.
.It Li p6-l2-dbus-busy
Count the number of cycles during which the L2 cache data bus was busy.
.It Li p6-l2-dbus-busy-rd
Count the number of cycles during which the L2 cache data bus was busy
transferring read data from L2 to the processor.
.It Li p6-l2-ifetch Op Li ,umask= Ns Ar qualifier
Count the number of L2 instruction fetches.
An additional qualifier may be specified and comprises a list of the following
keywords separated by
.Ql +
characters:
.Pp
.Bl -tag -width indent -compact
.It Li e
Count operations affecting E (exclusive) state lines.
.It Li i
Count operations affecting I (invalid) state lines.
.It Li m
Count operations affecting M (modified) state lines.
.It Li s
Count operations affecting S (shared) state lines.
.El
.Pp
The default is to count operations affecting all (MESI) state lines.
.It Li p6-l2-ld Op Li ,umask= Ns Ar qualifier
Count the number of L2 data loads.
An additional qualifier may be specified and comprises a list of the following
keywords separated by
.Ql +
characters:
.Pp
.Bl -tag -width indent -compact
.It Li both
.Pq Tn "Pentium M"
Count both hardware-prefetched lines and non-hardware-prefetched lines.
.It Li e
Count operations affecting E (exclusive) state lines.
.It Li hw
.Pq Tn "Pentium M"
Count hardware-prefetched lines only.
.It Li i
Count operations affecting I (invalid) state lines.
.It Li m
Count operations affecting M (modified) state lines.
.It Li nonhw
.Pq Tn "Pentium M"
Exclude hardware-prefetched lines.
.It Li s
Count operations affecting S (shared) state lines.
.El
.Pp
The default on processors other than
.Tn "Pentium M"
processors is to count operations affecting all (MESI) state lines.
The default on
.Tn "Pentium M"
processors is to count both hardware-prefetched and
non-hardware-prefetch operations on all (MESI) state lines.
.Pq Errata
This event is affected by processor errata E53.
.It Li p6-l2-lines-in Op Li ,umask= Ns Ar qualifier
Count the number of L2 lines allocated.
An additional qualifier may be specified and comprises a list of the following
keywords separated by
.Ql +
characters:
.Pp
.Bl -tag -width indent -compact
.It Li both
.Pq Tn "Pentium M"
Count both hardware-prefetched lines and non-hardware-prefetched lines.
.It Li e
Count operations affecting E (exclusive) state lines.
.It Li hw
.Pq Tn "Pentium M"
Count hardware-prefetched lines only.
.It Li i
Count operations affecting I (invalid) state lines.
.It Li m
Count operations affecting M (modified) state lines.
.It Li nonhw
.Pq Tn "Pentium M"
Exclude hardware-prefetched lines.
.It Li s
Count operations affecting S (shared) state lines.
.El
.Pp
The default on processors other than
.Tn "Pentium M"
processors is to count operations affecting all (MESI) state lines.
The default on
.Tn "Pentium M"
processors is to count both hardware-prefetched and
non-hardware-prefetch operations on all (MESI) state lines.
.Pq Errata
This event is affected by processor errata E45.
.It Li p6-l2-lines-out Op Li ,umask= Ns Ar qualifier
Count the number of L2 lines evicted.
An additional qualifier may be specified and comprises a list of the following
keywords separated by
.Ql +
characters:
.Pp
.Bl -tag -width indent -compact
.It Li both
.Pq Tn "Pentium M"
Count both hardware-prefetched lines and non-hardware-prefetched lines.
.It Li e
Count operations affecting E (exclusive) state lines.
.It Li hw
.Pq Tn "Pentium M"
Count hardware-prefetched lines only.
.It Li i
Count operations affecting I (invalid) state lines.
.It Li m
Count operations affecting M (modified) state lines.
.It Li nonhw
.Pq Tn "Pentium M" only
Exclude hardware-prefetched lines.
.It Li s
Count operations affecting S (shared) state lines.
.El
.Pp
The default on processors other than
.Tn "Pentium M"
processors is to count operations affecting all (MESI) state lines.
The default on
.Tn "Pentium M"
processors is to count both hardware-prefetched and
non-hardware-prefetch operations on all (MESI) state lines.
.Pq Errata
This event is affected by processor errata E45.
.It Li p6-l2-m-lines-inm
Count the number of modified lines allocated in L2 cache.
.It Li p6-l2-m-lines-outm Op Li ,umask= Ns Ar qualifier
Count the number of L2 M-state lines evicted.
.Pp
.Pq Tn "Pentium M"
On these processors an additional qualifier may be specified and
comprises a list of the following keywords separated by
.Ql +
characters:
.Pp
.Bl -tag -width indent -compact
.It Li both
Count both hardware-prefetched lines and non-hardware-prefetched lines.
.It Li hw
Count hardware-prefetched lines only.
.It Li nonhw
Exclude hardware-prefetched lines.
.El
.Pp
The default is to count both hardware-prefetched and
non-hardware-prefetch operations.
.Pq Errata
This event is affected by processor errata E53.
.It Li p6-l2-rqsts Op Li ,umask= Ns Ar qualifier
Count the total number of L2 requests.
An additional qualifier may be specified and comprises a list of the following
keywords separated by
.Ql +
characters:
.Pp
.Bl -tag -width indent -compact
.It Li e
Count operations affecting E (exclusive) state lines.
.It Li i
Count operations affecting I (invalid) state lines.
.It Li m
Count operations affecting M (modified) state lines.
.It Li s
Count operations affecting S (shared) state lines.
.El
.Pp
The default is to count operations affecting all (MESI) state lines.
.It Li p6-l2-st
Count the number of L2 data stores.
An additional qualifier may be specified and comprises a list of the following
keywords separated by
.Ql +
characters:
.Pp
.Bl -tag -width indent -compact
.It Li e
Count operations affecting E (exclusive) state lines.
.It Li i
Count operations affecting I (invalid) state lines.
.It Li m
Count operations affecting M (modified) state lines.
.It Li s
Count operations affecting S (shared) state lines.
.El
.Pp
The default is to count operations affecting all (MESI) state lines.
.It Li p6-ld-blocks
Count the number of load operations delayed due to store buffer blocks.
.It Li p6-misalign-mem-ref
Count the number of misaligned data memory references (crossing a 64
bit boundary).
.It Li p6-mmx-assist
.Pq Tn "Pentium II" , Tn "Pentium III"
Count the number of MMX assists executed.
.It Li p6-mmx-instr-exec
.Pq Tn Celeron , Tn "Pentium II"
Count the number of MMX instructions executed, except MOVQ and MOVD
stores from register to memory.
.It Li p6-mmx-instr-ret
.Pq Tn "Pentium II"
Count the number of MMX instructions retired.
.It Li p6-mmx-instr-type-exec Op Li ,umask= Ns Ar qualifier
.Pq Tn "Pentium II" , Tn "Pentium III"
Count the number of MMX instructions executed.
An additional qualifier may be specified and comprises a list of
the following keywords separated by
.Ql +
characters:
.Pp
.Bl -tag -width indent -compact
.It Li pack
Count MMX pack operation instructions.
.It Li packed-arithmetic
Count MMX packed arithmetic instructions.
.It Li packed-logical
Count MMX packed logical instructions.
.It Li packed-multiply
Count MMX packed multiply instructions.
.It Li packed-shift
Count MMX packed shift instructions.
.It Li unpack
Count MMX unpack operation instructions.
.El
.Pp
The default is to count all operations.
.It Li p6-mmx-sat-instr-exec
.Pq Tn "Pentium II" , Tn "Pentium III"
Count the number of MMX saturating instructions executed.
.It Li p6-mmx-uops-exec
.Pq Tn "Pentium II" , Tn "Pentium III"
Count the number of MMX micro-ops executed.
.It Li p6-mul
Count the number of integer and floating-point multiplies, including
speculative multiplies.
This event is only allocated on counter 1.
.It Li p6-partial-rat-stalls
Count the number of cycles or events for partial stalls.
.It Li p6-resource-stalls
Count the number of cycles there was a resource related stall of any kind.
.It Li p6-ret-seg-renames
.Pq Tn "Pentium II" , Tn "Pentium III"
Count the number of segment register rename events retired.
.It Li p6-sb-drains
Count the number of cycles the store buffer is draining.
.It Li p6-seg-reg-renames Op Li ,umask= Ns Ar qualifier
.Pq Tn "Pentium II" , Tn "Pentium III"
Count the number of segment register renames.
An additional qualifier may be specified, and comprises a list of the
following keywords separated by
.Ql +
characters:
.Pp
.Bl -tag -width indent -compact
.It Li ds
Count renames for segment register DS.
.It Li es
Count renames for segment register ES.
.It Li fs
Count renames for segment register FS.
.It Li gs
Count renames for segment register GS.
.El
.Pp
The default is to count operations affecting all segment registers.
.It Li p6-seg-rename-stalls
.Pq Tn "Pentium II" , Tn "Pentium III"
Count the number of segment register renaming stalls.
An additional qualifier may be specified, and comprises a list of the
following keywords separated by
.Ql +
characters:
.Pp
.Bl -tag -width indent -compact
.It Li ds
Count stalls for segment register DS.
.It Li es
Count stalls for segment register ES.
.It Li fs
Count stalls for segment register FS.
.It Li gs
Count stalls for segment register GS.
.El
.Pp
The default is to count operations affecting all the segment registers.
.It Li p6-segment-reg-loads
Count the number of segment register loads.
.It Li p6-uops-retired
Count the number of micro-ops retired.
.El
.Ss Intel P4 PMCS
Intel P4 PMCs are present in Intel
.Tn "Pentium 4"
and
.Tn Xeon
processors.
These PMCs are documented in
.Rs
.%B "IA-32 Intel(R) Architecture Software Developer's Manual"
.%T "Volume 3: System Programming Guide"
.%N "Order Number 245472-012"
.%D 2003
.%Q "Intel Corporation"
.Re
Further information about using these PMCs may be found in
.Rs
.%B "IA-32 Intel(R) Architecture Optimization Guide"
.%D 2003
.%N "Order Number 248966-009"
.%Q "Intel Corporation"
.Re
Some of these events are affected by processor errata described in
.Rs
.%B "Intel(R) Pentium(R) 4 Processor Specification Update"
.%N "Document Number: 249199-059"
.%D "April 2005"
.%Q "Intel Corporation"
.Re
.Pp
Event specifiers for Intel P4 PMCs can have the following common
qualifiers:
.Bl -tag -width indent
.It Li active= Ns Ar choice
(On P4 HTT CPUs) Filter event counting based on which logical
processors are active.
The allowed values of
.Ar choice
are:
.Pp
.Bl -tag -width indent -compact
.It Li any
Count when either logical processor is active.
.It Li both
Count when both logical processors are active.
.It Li none
Count only when neither logical processor is active.
.It Li single
Count only when one logical processor is active.
.El
.Pp
The default is
.Dq Li both .
.It Li cascade
Configure the PMC to cascade onto its partner.
See
.Sx "Cascading P4 PMCs"
below for more information.
.It Li edge
Configure the counter to count false to true transitions of the threshold
comparision output.
This qualifier only takes effect if a threshold qualifier has also been
specified.
.It Li complement
Configure the counter to increment only when the event count seen is
less than the threshold qualifier value specified.
.It Li mask= Ns Ar qualifier
Many event specifiers for Intel P4 PMCs need to be additionally
qualified using a mask qualifier.
The allowed syntax for these qualifiers is event specific and is
described along with the events.
.It Li os
Configure the PMC to count when the CPL of the processor is 0.
.It Li precise
Select precise event based sampling.
Precise sampling is supported by the hardware for a limited set of
events.
.It Li tag= Ns Ar value
Configure the PMC to tag the internal uop selected by the other
fields in this event specifier with value
.Ar value .
This feature is used when cascading PMCs.
.It Li threshold= Ns Ar value
Configure the PMC to increment only when the event counts seen are
greater than the specified threshold value
.Ar value .
.It Li usr
Configure the PMC to count when the CPL of the processor is 1, 2 or 3.
.El
.Pp
If neither of the
.Dq Li os
or
.Dq Li usr
qualifiers are specified, the default is to enable both.
.Pp
On Intel Pentium 4 processors with HTT, events are
divided into two classes:
.Pp
.Bl -tag -width indent -compact
.It "TS Events"
are those where hardware can differentiate between events
generated on one logical processor from those generated on the
other.
.It "TI Events"
are those where hardware cannot differentiate between events
generated by multiple logical processors in a package.
.El
.Pp
Only TS events are allowed for use with process-mode PMCs on
Pentium-4/HTT CPUs.
.Pp
The event specifiers supported by Intel P4 PMCs are:
.Pp
.Bl -tag -width indent
.It Li p4-128bit-mmx-uop Op Li ,mask= Ns Ar flags
.Pq "TI event"
Count integer SIMD SSE2 instructions that operate on 128 bit SIMD
operands.
Qualifier
.Ar flags
can take the following value (which is also the default):
.Pp
.Bl -tag -width indent -compact
.It Li all
Count all uops operating on 128 bit SIMD integer operands in memory or
XMM register.
.El
.Pp
If an instruction contains more than one 128 bit MMX uop, then each
uop will be counted.
.It Li p4-64bit-mmx-uop Op Li ,mask= Ns Ar flags
.Pq "TI event"
Count MMX instructions that operate on 64 bit SIMD operands.
Qualifier
.Ar flags
can take the following value (which is also the default):
.Pp
.Bl -tag -width indent -compact
.It Li all
Count all uops operating on 64 bit SIMD integer operands in memory or
in MMX registers.
.El
.Pp
If an instruction contains more than one 64 bit MMX uop, then each
uop will be counted.
.It Li p4-b2b-cycles
.Pq "TI event"
Count back-to-back bus cycles.
Further documentation for this event is unavailable.
.It Li p4-bnr
.Pq "TI event"
Count bus-not-ready conditions.
Further documentation for this event is unavailable.
.It Li p4-bpu-fetch-request Op Li ,mask= Ns Ar qualifier
.Pq "TS event"
Count instruction fetch requests qualified by additional
flags specified in
.Ar qualifier .
At this point only one flag is supported:
.Pp
.Bl -tag -width indent -compact
.It Li tcmiss
Count trace cache lookup misses.
.El
.Pp
The default qualifier is also
.Dq Li mask=tcmiss .
.It Li p4-branch-retired Op Li ,mask= Ns Ar flags
.Pq "TS event"
Counts retired branches.
Qualifier
.Ar flags
is a list of the following
.Ql +
separated strings:
.Pp
.Bl -tag -width indent -compact
.It Li mmnp
Count branches not-taken and predicted.
.It Li mmnm
Count branches not-taken and mis-predicted.
.It Li mmtp
Count branches taken and predicted.
.It Li mmtm
Count branches taken and mis-predicted.
.El
.Pp
The default qualifier counts all four kinds of branches.
.It Li p4-bsq-active-entries Op Li ,mask= Ns Ar qualifier
.Pq "TS event"
Count the number of entries (clipped at 15) currently active in the
BSQ.
Qualifier
.Ar qualifier
is a
.Ql +
separated set of the following flags:
.Pp
.Bl -tag -width indent -compact
.It Li req-type0 , Li req-type1
Forms a 2-bit number used to select the request type encoding:
.Pp
.Bl -tag -width indent -compact
.It Li 0
reads excluding read invalidate
.It Li 1
read invalidates
.It Li 2
writes other than writebacks
.It Li 3
writebacks
.El
.Pp
Bit
.Dq Li req-type1
is the MSB for this two bit number.
.It Li req-len0 , Li req-len1
Forms a two-bit number that specifies the request length encoding:
.Pp
.Bl -tag -width indent -compact
.It Li 0
0 chunks
.It Li 1
1 chunk
.It Li 3
8 chunks
.El
.Pp
Bit
.Dq Li req-len1
is the MSB for this two bit number.
.It Li req-io-type
Count requests that are input or output requests.
.It Li req-lock-type
Count requests that lock the bus.
.It Li req-lock-cache
Count requests that lock the cache.
.It Li req-split-type
Count requests that is a bus 8-byte chunk that is split across an
8-byte boundary.
.It Li req-dem-type
Count requests that are demand (not prefetches) if set.
Count requests that are prefetches if not set.
.It Li req-ord-type
Count requests that are ordered.
.It Li mem-type0 , Li mem-type1 , Li mem-type2
Forms a 3-bit number that specifies a memory type encoding:
.Pp
.Bl -tag -width indent -compact
.It Li 0
UC
.It Li 1
USWC
.It Li 4
WT
.It Li 5
WP
.It Li 6
WB
.El
.Pp
Bit
.Dq Li mem-type2
is the MSB of this 3-bit number.
.El
.Pp
The default qualifier has all the above bits set.
.Pp
Edge triggering using the
.Dq Li edge
qualifier should not be used with this event when counting cycles.
.It Li p4-bsq-allocation Op Li ,mask= Ns Ar qualifier
.Pq "TS event"
Count allocations in the bus sequence unit according to the flags
specified in
.Ar qualifier ,
which is a
.Ql +
separated set of the following flags:
.Pp
.Bl -tag -width indent -compact
.It Li req-type0 , Li req-type1
Forms a 2-bit number used to select the request type encoding:
.Pp
.Bl -tag -width indent -compact
.It Li 0
reads excluding read invalidate
.It Li 1
read invalidates
.It Li 2
writes other than writebacks
.It Li 3
writebacks
.El
.Pp
Bit
.Dq Li req-type1
is the MSB for this two bit number.
.It Li req-len0 , Li req-len1
Forms a two-bit number that specifies the request length encoding:
.Pp
.Bl -tag -width indent -compact
.It Li 0
0 chunks
.It Li 1
1 chunk
.It Li 3
8 chunks
.El
.Pp
Bit
.Dq Li req-len1
is the MSB for this two bit number.
.It Li req-io-type
Count requests that are input or output requests.
.It Li req-lock-type
Count requests that lock the bus.
.It Li req-lock-cache
Count requests that lock the cache.
.It Li req-split-type
Count requests that is a bus 8-byte chunk that is split across an
8-byte boundary.
.It Li req-dem-type
Count requests that are demand (not prefetches) if set.
Count requests that are prefetches if not set.
.It Li req-ord-type
Count requests that are ordered.
.It Li mem-type0 , Li mem-type1 , Li mem-type2
Forms a 3-bit number that specifies a memory type encoding:
.Pp
.Bl -tag -width indent -compact
.It Li 0
UC
.It Li 1
USWC
.It Li 4
WT
.It Li 5
WP
.It Li 6
WB
.El
.Pp
Bit
.Dq Li mem-type2
is the MSB of this 3-bit number.
.El
.Pp
The default qualifier has all the above bits set.
.Pp
This event is usually used along with the
.Dq Li edge
qualifier to avoid multiple counting.
.It Li p4-bsq-cache-reference Op Li ,mask= Ns Ar qualifier
.Pq "TS event"
Count cache references as seen by the bus unit (2nd or 3rd level
cache references).
Qualifier
.Ar qualifier
is a
.Ql +
separated list of the following keywords:
.Pp
.Bl -tag -width indent -compact
.It Li rd-2ndl-hits
Count 2nd level cache hits in the shared state.
.It Li rd-2ndl-hite
Count 2nd level cache hits in the exclusive state.
.It Li rd-2ndl-hitm
Count 2nd level cache hits in the modified state.
.It Li rd-3rdl-hits
Count 3rd level cache hits in the shared state.
.It Li rd-3rdl-hite
Count 3rd level cache hits in the exclusive state.
.It Li rd-3rdl-hitm
Count 3rd level cache hits in the modified state.
.It Li rd-2ndl-miss
Count 2nd level cache misses.
.It Li rd-3rdl-miss
Count 3rd level cache misses.
.It Li wr-2ndl-miss
Count write-back lookups from the data access cache that miss the 2nd
level cache.
.El
.Pp
The default is to count all the above events.
.It Li p4-execution-event Op Li ,mask= Ns Ar flags
.Pq "TS event"
Count the retirement of tagged uops selected through the execution
tagging mechanism.
Qualifier
.Ar flags
can contain the following strings separated by
.Ql +
characters:
.Pp
.Bl -tag -width indent -compact
.It Li nbogus0 , Li nbogus1 , Li nbogus2 , Li nbogus3
The marked uops are not bogus.
.It Li bogus0 , Li bogus1 , Li bogus2 , Li bogus3
The marked uops are bogus.
.El
.Pp
This event requires additional (upstream) events to be allocated to
perform the desired uop tagging.
The default is to set all the above flags.
This event can be used for precise event based sampling.
.It Li p4-front-end-event Op Li ,mask= Ns Ar flags
.Pq "TS event"
Count the retirement of tagged uops selected through the front-end
tagging mechanism.
Qualifier
.Ar flags
can contain the following strings separated by
.Ql +
characters:
.Pp
.Bl -tag -width indent -compact
.It Li nbogus
The marked uops are not bogus.
.It Li bogus
The marked uops are bogus.
.El
.Pp
This event requires additional (upstream) events to be allocated to
perform the desired uop tagging.
The default is to select both kinds of events.
This event can be used for precise event based sampling.
.It Li p4-fsb-data-activity Op Li ,mask= Ns Ar flags
.Pq "TI event"
Count each DBSY or DRDY event selected by qualifier
.Ar flags .
Qualifier
.Ar flags
is a
.Ql +
separated set of the following flags:
.Pp
.Bl -tag -width indent -compact
.It Li drdy-drv
Count when this processor is driving data onto the bus.
.It Li drdy-own
Count when this processor is reading data from the bus.
.It Li drdy-other
Count when data is on the bus but not being sampled by this processor.
.It Li dbsy-drv
Count when this processor reserves the bus for use in the next cycle
in order to drive data.
.It Li dbsy-own
Count when some agent reserves the bus for use in the next bus cycle
to drive data that this processor will sample.
.It Li dbsy-other
Count when some agent reserves the bus for use in the next bus cycle
to drive data that this processor will not sample.
.El
.Pp
Flags
.Dq Li drdy-own
and
.Dq Li drdy-other
are mutually exclusive.
Flags
.Dq Li dbsy-own
and
.Dq Li dbsy-other
are mutually exclusive.
The default value for
.Ar qualifier
is
.Dq Li drdy-drv+drdy-own+dbsy-drv+dbsy-own .
.It Li p4-global-power-events Op Li ,mask= Ns Ar flags
.Pq "TS event"
Count cycles during which the processor is not stopped.
Qualifier
.Ar flags
can take the following value (which is also the default):
.Pp
.Bl -tag -width indent -compact
.It Li running
Count cycles when the processor is active.
.El
.Pp
.It Li p4-instr-retired Op Li ,mask= Ns Ar flags
.Pq "TS event"
Count instructions retired during a clock cycle.
Qualifer
.Ar flags
comprises of the following strings separated by
.Ql +
characters:
.Pp
.Bl -tag -width indent -compact
.It Li nbogusntag
Count non-bogus instructions that are not tagged.
.It Li nbogustag
Count non-bogus instructions that are tagged.
.It Li bogusntag
Count bogus instructions that are not tagged.
.It Li bogustag
Count bogus instructions that are tagged.
.El
.Pp
The default qualifier counts all the above kinds of instructions.
.It Li p4-ioq-active-entries Xo
.Op Li ,mask= Ns Ar qualifier
.Op Li ,busreqtype= Ns Ar req-type
.Xc
.Pq "TS event"
Count the number of entries (clipped at 15) in the IOQ that are
active.
The event masks are specified by qualifier
.Ar qualifier
and
.Ar req-type .
.Pp
Qualifier
.Ar qualifier
is a
.Ql +
separated set of the following flags:
.Pp
.Bl -tag -width indent -compact
.It Li all-read
Count read entries.
.It Li all-write
Count write entries.
.It Li mem-uc
Count entries accessing uncacheable memory.
.It Li mem-wc
Count entries accessing write-combining memory.
.It Li mem-wt
Count entries accessing write-through memory.
.It Li mem-wp
Count entries accessing write-protected memory
.It Li mem-wb
Count entries accessing write-back memory.
.It Li own
Count store requests driven by the processor (i.e., not by other
processors or by DMA).
.It Li other
Count store requests driven by other processors or by DMA.
.It Li prefetch
Include hardware and software prefetch requests in the count.
.El
.Pp
The default value for
.Ar qualifier
is to enable all the above flags.
.Pp
The
.Ar req-type
qualifier is a 5-bit number can be additionally used to select a
specific bus request type.
The default is 0.
.Pp
The
.Dq Li edge
qualifier should not be used when counting cycles with this event.
The exact behaviour of this event depends on the processor revision.
.It Li p4-ioq-allocation Xo
.Op Li ,mask= Ns Ar qualifier
.Op Li ,busreqtype= Ns Ar req-type
.Xc
.Pq "TS event"
Count various types of transactions on the bus matching the flags set
in
.Ar qualifier
and
.Ar req-type .
.Pp
Qualifier
.Ar qualifier
is a
.Ql +
separated set of the following flags:
.Pp
.Bl -tag -width indent -compact
.It Li all-read
Count read entries.
.It Li all-write
Count write entries.
.It Li mem-uc
Count entries accessing uncacheable memory.
.It Li mem-wc
Count entries accessing write-combining memory.
.It Li mem-wt
Count entries accessing write-through memory.
.It Li mem-wp
Count entries accessing write-protected memory
.It Li mem-wb
Count entries accessing write-back memory.
.It Li own
Count store requests driven by the processor (i.e., not by other
processors or by DMA).
.It Li other
Count store requests driven by other processors or by DMA.
.It Li prefetch
Include hardware and software prefetch requests in the count.
.El
.Pp
The default value for
.Ar qualifier
is to enable all the above flags.
.Pp
The
.Ar req-type
qualifier is a 5-bit number can be additionally used to select a
specific bus request type.
The default is 0.
.Pp
The
.Dq Li edge
qualifier is normally used with this event to prevent multiple
counting.
The exact behaviour of this event depends on the processor revision.
.It Li p4-itlb-reference Op mask= Ns Ar qualifier
.Pq "TS event"
Count translations using the intruction translation look-aside
buffer.
The
.Ar qualifier
argument is a list of the following strings separated by
.Ql +
characters.
.Pp
.Bl -tag -width indent -compact
.It Li hit
Count ITLB hits.
.It Li miss
Count ITLB misses.
.It Li hit-uc
Count uncacheable ITLB hits.
.El
.Pp
If no
.Ar qualifier
is specified the default is to count all the three kinds of ITLB
translations.
.It Li p4-load-port-replay Op Li ,mask= Ns Ar qualifier
.Pq "TS event"
Count replayed events at the load port.
Qualifier
.Ar qualifier
can take on one value:
.Pp
.Bl -tag -width indent -compact
.It Li split-ld
Count split loads.
.El
.Pp
The default value for
.Ar qualifier
is
.Dq Li split-ld .
.It Li p4-mispred-branch-retired Op Li ,mask= Ns Ar flags
.Pq "TS event"
Count mispredicted IA-32 branch instructions.
Qualifier
.Ar flags
can take the following value (which is also the default):
.Pp
.Bl -tag -width indent -compact
.It Li nbogus
Count non-bogus retired branch instructions.
.El
.It Li p4-machine-clear Op Li ,mask= Ns Ar flags
.Pq "TS event"
Count the number of pipeline clears seen by the processor.
Qualifer
.Ar flags
is a list of the following strings separated by
.Ql +
characters:
.Pp
.Bl -tag -width indent -compact
.It Li clear
Count for a portion of the many cycles when the machine is being
cleared for any reason.
.It Li moclear
Count machine clears due to memory ordering issues.
.It Li smclear
Count machine clears due to self-modifying code.
.El
.Pp
Use qualifier
.Dq Li edge
to get a count of occurrences of machine clears.
The default qualifier is
.Dq Li clear .
.It Li p4-memory-cancel Op Li ,mask= Ns Ar event-list
.Pq "TS event"
Count the cancelling of various kinds of requests in the data cache
address control unit of the CPU.
The qualifier
.Ar event-list
is a list of the following strings separated by
.Ql +
characters:
.Pp
.Bl -tag -width indent -compact
.It Li st-rb-full
Requests cancelled because no store request buffer was available.
.It Li 64k-conf
Requests that conflict due to 64K aliasing.
.El
.Pp
If
.Ar event-list
is not specified, then the default is to count both kinds of events.
.It Li p4-memory-complete Op Li ,mask= Ns Ar event-list
.Pq "TS event"
Count the completion of load split, store split, uncacheable split and
uncacheable load operations selected by qualifier
.Ar event-list .
The qualifier
.Ar event-list
is a
.Ql +
separated list of the following flags:
.Pp
.Bl -tag -width indent -compact
.It Li lsc
Count load splits completed, excluding loads from uncacheable or
write-combining areas.
.It Li ssc
Count any split stores completed.
.El
.Pp
The default is to count both kinds of operations.
.It Li p4-mob-load-replay Op Li ,mask= Ns Ar qualifier
.Pq "TS event"
Count load replays triggered by the memory order buffer.
Qualifier
.Ar qualifier
can be a
.Ql +
separated list of the following flags:
.Pp
.Bl -tag -width indent -compact
.It Li no-sta
Count replays because of unknown store addresses.
.It Li no-std
Count replays because of unknown store data.
.It Li partial-data
Count replays because of partially overlapped data accesses between
load and store operations.
.It Li unalgn-addr
Count replays because of mismatches in the lower 4 bits of load and
store operations.
.El
.Pp
The default qualifier is
.Ar no-sta+no-std+partial-data+unalgn-addr .
.It Li p4-packed-dp-uop Op Li ,mask= Ns Ar flags
.Pq "TI event"
Count packed double-precision uops.
Qualifier
.Ar flags
can take the following value (which is also the default):
.Pp
.Bl -tag -width indent -compact
.It Li all
Count all uops operating on packed double-precision operands.
.El
.It Li p4-packed-sp-uop Op Li ,mask= Ns Ar flags
.Pq "TI event"
Count packed single-precision uops.
Qualifier
.Ar flags
can take the following value (which is also the default):
.Pp
.Bl -tag -width indent -compact
.It Li all
Count all uops operating on packed single-precision operands.
.El
.It Li p4-page-walk-type Op Li ,mask= Ns Ar qualifier
.Pq "TI event"
Count page walks performed by the page miss handler.
Qualifier
.Ar qualifier
can be a
.Ql +
separated list of the following keywords:
.Pp
.Bl -tag -width indent -compact
.It Li dtmiss
Count page walks for data TLB misses.
.It Li itmiss
Count page walks for instruction TLB misses.
.El
.Pp
The default value for
.Ar qualifier
is
.Dq Li dtmiss+itmiss .
.It Li p4-replay-event Op Li ,mask= Ns Ar flags
.Pq "TS event"
Count the retirement of tagged uops selected through the replay
tagging mechanism.
Qualifier
.Ar flags
contains a
.Ql +
separated set of the following strings:
.Pp
.Bl -tag -width indent -compact
.It Li nbogus
The marked uops are not bogus.
.It Li bogus
The marked uops are bogus.
.El
.Pp
This event requires additional (upstream) events to be allocated to
perform the desired uop tagging.
The default qualifier counts both kinds of uops.
This event can be used for precise event based sampling.
.It Li p4-resource-stall Op Li ,mask= Ns Ar flags
.Pq "TS event"
Count the occurrence or latency of stalls in the allocator.
Qualifier
.Ar flags
can take the following value (which is also the default):
.Pp
.Bl -tag -width indent -compact
.It Li sbfull
A stall due to the lack of store buffers.
.El
.It Li p4-response
.Pq "TI event"
Count different types of responses.
Further documentation on this event is not available.
.It Li p4-retired-branch-type Op Li ,mask= Ns Ar flags
.Pq "TS event"
Count branches retired.
Qualifier
.Ar flags
contains a
.Ql +
separated list of strings:
.Pp
.Bl -tag -width indent -compact
.It Li conditional
Count conditional jumps.
.It Li call
Count direct and indirect call branches.
.It Li return
Count return branches.
.It Li indirect
Count returns, indirect calls or indirect jumps.
.El
.Pp
The default qualifier counts all the above branch types.
.It Li p4-retired-mispred-branch-type Op Li ,mask= Ns Ar flags
.Pq "TS event"
Count mispredicted branches retired.
Qualifier
.Ar flags
contains a
.Ql +
separated list of strings:
.Pp
.Bl -tag -width indent -compact
.It Li conditional
Count conditional jumps.
.It Li call
Count indirect call branches.
.It Li return
Count return branches.
.It Li indirect
Count returns, indirect calls or indirect jumps.
.El
.Pp
The default qualifier counts all the above branch types.
.It Li p4-scalar-dp-uop Op Li ,mask= Ns Ar flags
.Pq "TI event"
Count the number of scalar double-precision uops.
Qualifier
.Ar flags
can take the following value (which is also the default):
.Pp
.Bl -tag -width indent -compact
.It Li all
Count the number of scalar double-precision uops.
.El
.It Li p4-scalar-sp-uop Op Li ,mask= Ns Ar flags
.Pq "TI event"
Count the number of scalar single-precision uops.
Qualifier
.Ar flags
can take the following value (which is also the default):
.Pp
.Bl -tag -width indent -compact
.It Li all
Count all uops operating on scalar single-precision operands.
.El
.It Li p4-snoop
.Pq "TI event"
Count snoop traffic.
Further documentation on this event is not available.
.It Li p4-sse-input-assist Op Li ,mask= Ns Ar flags
.Pq "TI event"
Count the number of times an assist is required to handle problems
with the operands for SSE and SSE2 operations.
Qualifier
.Ar flags
can take the following value (which is also the default):
.Pp
.Bl -tag -width indent -compact
.It Li all
Count assists for all SSE and SSE2 uops.
.El
.It Li p4-store-port-replay Op Li ,mask= Ns Ar qualifier
.Pq "TS event"
Count events replayed at the store port.
Qualifier
.Ar qualifier
can take on one value:
.Pp
.Bl -tag -width indent -compact
.It Li split-st
Count split stores.
.El
.Pp
The default value for
.Ar qualifier
is
.Dq Li split-st .
.It Li p4-tc-deliver-mode Op Li ,mask= Ns Ar qualifier
.Pq "TI event"
Count the duration in cycles of operating modes of the trace cache and
decode engine.
The desired operating mode is selected by
.Ar qualifier ,
which is a list of the following strings separated by
.Ql +
characters:
.Pp
.Bl -tag -width indent -compact
.It Li DD
Both logical processors are in deliver mode.
.It Li DB
Logical processor 0 is in deliver mode while logical processor 1 is in
build mode.
.It Li DI
Logical processor 0 is in deliver mode while logical processor 1 is
halted, or in machine clear, or transitioning to a long microcode
flow.
.It Li BD
Logical processor 0 is in build mode while logical processor 1 is in
deliver mode.
.It Li BB
Both logical processors are in build mode.
.It Li BI
Logical processor 0 is in build mode while logical processor 1 is
halted, or in machine clear or transitioning to a long microcode
flow.
.It Li ID
Logical processor 0 is halted, or in machine clear or transitioning to
a long microcode flow while logical processor 1 is in deliver mode.
.It Li IB
Logical processor 0 is halted, or in machine clear or transitioning to
a long microcode flow while logical processor 1 is in build mode.
.El
.Pp
If there is only one logical processor in the processor package then
the qualifier for logical processor 1 is ignored.
If no qualifier is specified, the default qualifier is
.Dq Li DD+DB+DI+BD+BB+BI+ID+IB .
.It Li p4-tc-ms-xfer Op Li ,mask= Ns Ar flags
.Pq "TI event"
Count the number of times uop delivery changed from the trace cache to
MS ROM.
Qualifier
.Ar flags
can take the following value (which is also the default):
.Pp
.Bl -tag -width indent -compact
.It Li cisc
Count TC to MS transfers.
.El
.It Li p4-uop-queue-writes Op Li ,mask= Ns Ar flags
.Pq "TS event"
Count the number of valid uops written to the uop queue.
Qualifier
.Ar flags
is a list of the following strings, separated by
.Ql +
characters:
.Pp
.Bl -tag -width indent -compact
.It Li from-tc-build
Count uops being written from the trace cache in build mode.
.It Li from-tc-deliver
Count uops being written from the trace cache in deliver mode.
.It Li from-rom
Count uops being written from microcode ROM.
.El
.Pp
The default qualifier counts all the above kinds of uops.
.It Li p4-uop-type Op Li ,mask= Ns Ar flags
.Pq "TS event"
This event is used in conjunction with the front-end at-retirement
mechanism to tag load and store uops.
Qualifer
.Ar flags
comprises the following strings separated by
.Ql +
characters:
.Pp
.Bl -tag -width indent -compact
.It Li tagloads
Mark uops that are load operations.
.It Li tagstores
Mark uops that are store operations.
.El
.Pp
The default qualifier counts both kinds of uops.
.It Li p4-uops-retired Op Li ,mask= Ns Ar flags
.Pq "TS event"
Count uops retired during a clock cycle.
Qualifier
.Ar flags
comprises the following strings separated by
.Ql +
characters:
.Pp
.Bl -tag -width indent -compact
.It Li nbogus
Count marked uops that are not bogus.
.It Li bogus
Count marked uops that are bogus.
.El
.Pp
The default qualifier counts both kinds of uops.
.It Li p4-wc-buffer Op Li ,mask= Ns Ar flags
.Pq "TI event"
Count write-combining buffer operations.
Qualifier
.Ar flags
contains the following strings separated by
.Ql +
characters:
.Pp
.Bl -tag -width indent -compact
.It Li wcb-evicts
WC buffer evictions due to any cause.
.It Li wcb-full-evict
WC buffer evictions due to no WC buffer being available.
.El
.Pp
The default qualifer counts both kinds of evictions.
.It Li p4-x87-assist Op Li ,mask= Ns Ar flags
.Pq "TS event"
Count the retirement of x87 instructions that required special
handling.
Qualifier
.Ar flags
contains the following strings separated by
.Ql +
characters:
.Pp
.Bl -tag -width indent -compact
.It Li fpsu
Count instructions that saw an FP stack underflow.
.It Li fpso
Count instructions that saw an FP stack overflow.
.It Li poao
Count instructions that saw an x87 output overflow.
.It Li poau
Count instructions that saw an x87 output underflow.
.It Li prea
Count instructions that needed an x87 input assist.
.El
.Pp
The default qualifier counts all the above types of instruction
retirements.
.It Li p4-x87-fp-uop Op Li ,mask= Ns Ar flags
.Pq "TI event"
Count x87 floating-point uops.
Qualifier
.Ar flags
can take the following value (which is also the default):
.Pp
.Bl -tag -width indent -compact
.It Li all
Count all x87 floating-point uops.
.El
.Pp
If an instruction contains more than one x87 floating-point uops, then
all x87 floating-point uops will be counted.
This event does not count x87 floating-point data movement operations.
.It Li p4-x87-simd-moves-uop Op Li ,mask= Ns Ar flags
.Pq "TI event"
Count each x87 FPU, MMX, SSE, or SSE2 uops that load data or store
data or perform register-to-register moves.
This event does not count integer move uops.
Qualifier
.Ar flags
may contain the following keywords separated by
.Ql +
characters:
.Pp
.Bl -tag -width indent -compact
.It Li allp0
Count all x87 and SIMD store and move uops.
.It Li allp2
Count all x87 and SIMD load uops.
.El
.Pp
The default is to count all uops.
.Pq Errata
This event may be affected by processor errata N43.
.El
.Ss "Cascading P4 PMCs"
PMC cascading support is currently poorly implemented.
While individual event counters may be allocated with a
.Dq Li cascade
qualifier, the current API does not offer the ability
to name and allocate all the resources needed for a
cascaded event counter pair in a single operation.
.Ss "Precise Event Based Sampling"
Support for precise event based sampling is currently
unimplemented.
.Sh COMPATIBILITY
The interface between the
.Nm pmc
library and the
.Xr hwpmc 4
driver is intended to be private to the implementation and may
change.
In order to ease forward compatibility with future versions of the
.Xr hwpmc 4
driver, applications are urged to dynamically link with the
.Nm pmc
library.
.Pp
The
.Nm pmc
API is
.Ud
.Sh SEE ALSO
.Xr pmclog 3 ,
.Xr hwpmc 4 ,
.Xr pmccontrol 8 ,
.Xr pmcstat 8
.Sh HISTORY
The
.Nm pmc
library first appeared in
.Fx 6.0 .
.Sh AUTHORS
The
.Lb libpmc
library was written by
.An "Joseph Koshy"
.Aq jkoshy@FreeBSD.org .