pmc_allocatepmc_attachpmc_capabilitiespmc_configure_logfilepmc_cpuinfopmc_detachpmc_disablepmc_enablepmc_event_names_of_classpmc_flush_logfilepmc_get_driver_statspmc_get_msrpmc_initpmc_name_of_capabilitypmc_name_of_classpmc_name_of_cputypepmc_name_of_eventpmc_name_of_modepmc_name_of_statepmc_ncpupmc_npmcpmc_pmcinfopmc_readpmc_releasepmc_rwpmc_setpmc_startpmc_stoppmc_widthpmc_writepmc_writelog
- programming API for using hardware performance monitoring counters
LIBRARY
Lb libpmc
SYNOPSIS
#include <pmc.h> int
Fo pmc_allocate
Fa const char *eventspecifier
Fa enum pmc_mode mode
Fa uint32_t flags
Fa int cpu
Fa pmc_id_t *pmcid
Fc Ft int
pmc_attach (pmc_id_t pmcid pid_t pid); int
pmc_capabilities (pmc_id_t pmc uint32_t *caps); int
pmc_configure_logfile (int fd); int
pmc_cpuinfo (const struct pmc_cpuinfo **cpu_info); int
pmc_detach (pmc_id_t pmcid pid_t pid); int
pmc_disable (int cpu int pmc); int
pmc_enable (int cpu int pmc); int
Fo pmc_event_names_of_class
Fa enum pmc_class cl
Fa const char ***eventnames
Fa int *nevents
Fc Ft int
pmc_flush_logfile (void); int
pmc_get_driver_stats (struct pmc_driverstats *gms); int
pmc_get_msr (pmc_id_t pmc uint32_t *msr); int
pmc_init (void); const char *
pmc_name_of_capability (enum pmc_caps pc); const char *
pmc_name_of_class (enum pmc_class pc); const char *
pmc_name_of_cputype (enum pmc_cputype ct); const char *
pmc_name_of_disposition (enum pmc_disp pd); const char *
pmc_name_of_event (enum pmc_event pe); const char *
pmc_name_of_mode (enum pmc_mode pm); const char *
pmc_name_of_state (enum pmc_state ps); int
pmc_ncpu (void); int
pmc_npmc (int cpu); int
pmc_pmcinfo (int cpu struct pmc_pmcinfo **pmc_info); int
pmc_read (pmc_id_t pmc pmc_value_t *value); int
pmc_release (pmc_id_t pmc); int
pmc_rw (pmc_id_t pmc pmc_value_t newvalue pmc_value_t *oldvaluep); int
pmc_set (pmc_id_t pmc pmc_value_t value); int
pmc_start (pmc_id_t pmc); int
pmc_stop (pmc_id_t pmc); int
pmc_write (pmc_id_t pmc pmc_value_t value); int
pmc_writelog (uint32_t userdata); int
pmc_width (pmc_id_t pmc uint32_t *width);
DESCRIPTION
These functions implement a high-level library for using the
system's hardware performance counters.
PMCs are allocated using
pmc_allocate (,);
released using
pmc_release ();
and read using
pmc_read (.);
Allocated PMCs may be started or stopped at any time using
pmc_start ();
and
pmc_stop ();
respectively.
An allocated PMC may be of
``global''
scope, meaning that the PMC measures system-wide events, or
``process-private''
scope, meaning that the PMC only counts hardware events when
the allocating process (or, optionally, its children)
are active.
PMCs may further be in
``counting mode''
or in
``sampling mode''
Sampling mode PMCs deliver an interrupt to the CPU after
a configured number of hardware events have been seen.
A process-private sampling mode PMC will cause its owner
process to get periodic
SIGPROF
interrupts, while a global sampling mode PMC is used to
do system-wide statistical sampling (see
hwpmc(4)).
The sampling rate desired of a sampling-mode PMC is set using
pmc_set (.);
Counting mode PMCs do not interrupt the CPU; their values
can be read using
pmc_read (.);
System-wide statistical sampling is configured by allocating
at least one sampling mode PMC with
global scope, and when a log file is configured using
pmc_configure_logfile (.);
The
hwpmc(4)
driver manages system-wide statistical sampling; for more
information please see
hwpmc(4).
Application Programming Interface
The function
pmc_init ();
initializes the
pmc
library.
This function must be called first, before any of the other
functions in the library.
The function
pmc_allocate ();
allocates a counter that counts the events named by
Fa eventspecifier ,
and writes the allocated counter ID to
Fa *pmcid .
Argument
Fa eventspecifier
comprises an PMC event name followed by an optional comma separated
list of keywords and qualifiers.
The allowed syntax for
Fa eventspecifier
is processor architecture specific and is listed in section
Sx EVENT SPECIFIERS
below.
The desired PMC mode is specified by
Fa mode ,
and any mode specific modifiers are specified using
Fa flags .
The
Fa cpu
argument is the value
PMC_CPU_ANY
or names the CPU the allocation is to be on.
Requesting a specific CPU only makes sense for global PMCs;
process-private PMC allocations should always specify
PMC_CPU_ANY
By default, a PMC configured in process-virtual counting mode is set up
to profile its owner process.
The function
pmc_attach ();
may be used to attach the PMC to a different process.
It
needs to be called before the counter is first started
with
pmc_start (.);
The function
pmc_detach ();
may be used to detach a PMC from a process it was attached to
using a prior call to
pmc_attach (.);
The function
pmc_release ();
releases a PMC previously allocated with
pmc_allocate (.);
This function call implicitly detaches the PMC from all its target
processes.
An allocated PMC may be started and stopped using
pmc_start ();
and
pmc_stop ();
respectively.
The current value of a PMC may be read with
pmc_read ();
and written using
pmc_write (,);
provided the underlying hardware supports these operations on
the allocated PMC.
The read and write operation may be combined using
pmc_rw (.);
The function
pmc_capabilities ();
sets argument
Fa caps
to a bitmask of capabilities supported by the PMC denoted by
argument
Fa pmc .
The function
pmc_width ();
sets argument
Fa width
to the width of the PMC denoted by argument
Fa pmc .
The
pmc_configure_logfile ();
function causes the
hwpmc(4)
driver to log performance data to file corresponding
to the process' file handle
Fa fd .
If argument
Fa fd
is -1, then any previously configured logging is reset
and all data queued to be written are discarded.
The
pmc_flush_logfile ();
function will send all data queued inside the
hwpmc(4)
driver to the configured log file before returning.
The
pmc_writelog ();
function will append a log entry containing the argument
Fa userdata
to the log file.
The function
pmc_set ();
configures a sampling PMC
Fa pmc
to interrupt every
Fa value
events.
For counting PMCs,
pmc_set ();
sets the initial value of the PMC to
Fa value .
The function
pmc_get_driver_statistics ();
copies a snapshot of the usage statistics maintained by
hwpmc(4)
into the memory area pointed to by argument
Fa gms .
Signal Handling Requirements
Applications using PMCs are required to handle the following signals:
SIGBUS
When the
hwpmc(4)
module is unloaded using
kldunload(8),
processes that have PMCs allocated to them will be sent a
SIGBUS
signal.
SIGIO
The
hwpmc(4)
driver will send a PMC owning process a
SIGIO
signal if:
If any process-mode PMC allocated by it loses all its
target processes.
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
pmc_flush_logfile (.);
Convenience Functions
The function
pmc_ncpu ();
returns the number of CPUs present in the system.
The function
pmc_npmc ();
returns the number of PMCs supported on CPU
Fa cpu .
The function
pmc_cpuinfo ();
sets argument
Fa cpu_info
to point to an internal structure with information about the system's CPUs.
The caller should not
free ();
this pointer value.
Function
pmc_pmcinfo ();
returns information about the current state of CPU
Fa cpu Ns 's
PMCs.
This function sets argument
Fa *pmc_info
to point to a memory area allocated with
calloc(3).
The caller is expected to
free ();
the area when done.
The functions
pmc_name_of_capability (,);
pmc_name_of_class (,);
pmc_name_of_cputype (,);
pmc_name_of_disposition (,);
pmc_name_of_event (,);
pmc_name_of_mode ();
and
pmc_name_of_state ();
are useful for code wanting to print error messages.
They return
Vt const char *
pointers to human-readable representations of their arguments.
These return values should not be freed using
free(3).
The function
pmc_event_names_of_class ();
returns a list of event names supported by a given PMC class
Fa cl .
On successful return, an array of
Vt const char *
pointers to the names of valid events supported by class
Fa cl
is allocated by the library using
malloc(3),
and a pointer to this array is returned in the location pointed to by
Fa eventnames .
The number of pointers allocated is returned in the location pointed
to by
Fa nevents .
Administration
Individual PMCs may be enabled or disabled on a given CPU using
pmc_enable ();
and
pmc_disable ();
respectively.
For these functions,
Fa cpu
is the CPU number, and
Fa pmc
is the index of the PMC to be operated on.
Only the super-user is allowed to enable and disable PMCs.
x86 Architecture Specific API
The
pmc_get_msr ();
function returns the processor model specific register number
associated with
Fa pmc .
Applications may use the x86
RDPMC
instruction to directly read the contents of the PMC.
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
hwpmc(4)
library defines machine independent aliases for commonly used
events.
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
pmc
library:
branches
Measure the number of branches retired.
branch-mispredicts
Measure the number of retired branches that were mispredicted.
cycles
Measure processor cycles.
This event is implemented using the processor's Time Stamp Counter
register.
dc-misses
Measure the number of data cache misses.
ic-misses
Measure the number of instruction cache misses.
instructions
Measure the number of instructions retired.
interrupts
Measure the number of interrupts seen.
unhalted-cycles
Measure the number of cycles the processor is not in a halted
or sleep state.
Time Stamp Counter (TSC)
The timestamp counter is a monotonically non-decreasing counter that
counts processor cycles.
In the i386 architecture, this counter may
be selected by requesting an event with event specifier
``tsc
''
The
``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
pmc_read ();
function, or by using the RDTSC instruction.
Event specifiers for AMD K7 PMCs can have the following optional
qualifiers:
count= value
Configure the counter to increment only if the number of configured
events measured in a cycle is greater than or equal to
value
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.
inv
Invert the sense of comparision when the
``count
''
qualifier is present, making the counter to increment when the
number of events per cycle is less than the value specified by
the
``count
''
qualifier.
os
Configure the PMC to count events happening at privilege level 0.
unitmask= mask
This qualifier is used to further qualify a select few events,
``k7-dc-refills-from-l2
''
``k7-dc-refills-from-system
''
and
``k7-dc-writebacks
''
Here
mask
is a string of the following characters optionally separated by
`+'
characters:
m
Count operations for lines in the
``Modified''
state.
o
Count operations for lines in the
``Owner''
state.
e
Count operations for lines in the
``Exclusive''
state.
s
Count operations for lines in the
``Shared''
state.
i
Count operations for lines in the
``Invalid''
state.
If no
``unitmask
''
qualifier is specified, the default is to count events for caches
lines in any of the above states.
usr
Configure the PMC to count events occurring at privilege levels 1, 2
or 3.
If neither of the
``os
''
or
``usr
''
qualifiers were specified, the default is to enable both.
The event specifiers supported on AMD K7 PMCs are:
k7-dc-accesses
Count data cache accesses.
k7-dc-misses
Count data cache misses.
k7-dc-refills-from-l2 [,unitmask= mask
]
Count data cache refills from L2 cache.
This event may be further qualified using the
``unitmask
''
qualifier.
k7-dc-refills-from-system [,unitmask= mask
]
Count data cache refills from system memory.
This event may be further qualified using the
``unitmask
''
qualifier.
k7-dc-writebacks [,unitmask= mask
]
Count data cache writebacks.
This event may be further qualified using the
``unitmask
''
qualifier.
k7-l1-dtlb-miss-and-l2-dtlb-hits
Count L1 DTLB misses and L2 DTLB hits.
k7-l1-and-l2-dtlb-misses
Count L1 and L2 DTLB misses.
k7-misaligned-references
Count misaligned data references.
k7-ic-fetches
Count instruction cache fetches.
k7-ic-misses
Count instruction cache misses.
k7-l1-itlb-misses
Count L1 ITLB misses that are L2 ITLB hits.
k7-l1-l2-itlb-misses
Count L1 (and L2) ITLB misses.
k7-retired-instructions
Count all retired instructions.
k7-retired-ops
Count retired ops.
k7-retired-branches
Count all retired branches (conditional, unconditional, exceptions
and interrupts).
k7-retired-branches-mispredicted
Count all misprediced retired branches.
k7-retired-taken-branches
Count retired taken branches.
k7-retired-taken-branches-mispredicted
Count mispredicted taken branches that were retired.
k7-retired-far-control-transfers
Count retired far control transfers.
k7-retired-resync-branches
Count retired resync branches (non control transfer branches).
k7-interrupts-masked-cycles
Count the number of cycles when the processor's
IF
flag was zero.
k7-interrupts-masked-while-pending-cycles
Count the number of cycles interrupts were masked while pending due
to the processor's
IF
flag being zero.
k7-hardware-interrupts
Count the number of taken hardware interrupts.
AMD (K8) PMCs These PMCs are present in the
AMD Athlon64
and
AMD Opteron
series of CPUs.
They are documented in:
"BIOS and Kernel Developer's Guide for the AMD Athlon(tm) 64 and AMD Opteron Processors"
"Publication No. 26094"
"April 2004"
"Advanced Micro Devices, Inc."
Event specifiers for AMD K8 PMCs can have the following optional
qualifiers:
count= value
Configure the counter to increment only if the number of configured
events measured in a cycle is greater than or equal to
value
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.
inv
Invert the sense of comparision when the
``count
''
qualifier is present, making the counter to increment when the
number of events per cycle is less than the value specified by
the
``count
''
qualifier.
mask= 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.
os
Configure the PMC to count events happening at privilege level 0.
usr
Configure the PMC to count events occurring at privilege levels 1, 2
or 3.
If neither of the
``os
''
or
``usr
''
qualifiers were specified, the default is to enable both.
The event specifiers supported on AMD K8 PMCs are:
k8-bu-cpu-clk-unhalted
Count the number of clock cycles when the CPU is not in the HLT or
STPCLK states.
k8-bu-fill-request-l2-miss [,mask= qualifier
]
Count fill requests that missed in the L2 cache.
This event may be further qualified using
qualifier
which is a
`+'
separated set of the following keywords:
dc-fill
Count data cache fill requests.
ic-fill
Count instruction cache fill requests.
tlb-reload
Count TLB reloads.
The default is to count all types of requests.
k8-bu-internal-l2-request [,mask= qualifier
]
Count internally generated requests to the L2 cache.
This event may be further qualified using
qualifier
which is a
`+'
separated set of the following keywords:
cancelled
Count cancelled requests.
dc-fill
Count data cache fill requests.
ic-fill
Count instruction cache fill requests.
tag-snoop
Count tag snoop requests.
tlb-reload
Count TLB reloads.
The default is to count all types of requests.
k8-dc-access
Count data cache accesses including microcode scratchpad accesses.
k8-dc-copyback [,mask= qualifier
]
Count data cache copyback operations.
This event may be further qualified using
qualifier
which is a
`+'
separated set of the following keywords:
exclusive
Count operations for lines in the
``exclusive''
state.
invalid
Count operations for lines in the
``invalid''
state.
modified
Count operations for lines in the
``modified''
state.
owner
Count operations for lines in the
``owner''
state.
shared
Count operations for lines in the
``shared''
state.
The default is to count operations for lines in all the
above states.
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
qualifier
which is a
`+'
separated set of the following keywords:
Count the number of dispatched prefetch instructions.
This event may be further qualified using
qualifier
which is a
`+'
separated set of the following keywords:
load
Count load operations.
nta
Count non-temporal operations.
store
Count store operations.
The default is to count all operations.
k8-dc-l1-dtlb-miss-and-l2-dtlb-hit
Count L1 DTLB misses that are L2 DTLB hits.
k8-dc-l1-dtlb-miss-and-l2-dtlb-miss
Count L1 DTLB misses that are also misses in the L2 DTLB.
Count microarchitectural early cancels of data cache accesses.
k8-dc-microarchitectural-late-cancel-of-an-access
Count microarchitectural late cancels of data cache accesses.
k8-dc-misaligned-data-reference
Count misaligned data references.
k8-dc-miss
Count data cache misses.
k8-dc-one-bit-ecc-error [,mask= qualifier
]
Count one bit ECC errors found by the scrubber.
This event may be further qualified using
qualifier
which is a
`+'
separated set of the following keywords:
scrubber
Count scrubber detected errors.
piggyback
Count piggyback scrubber errors.
The default is to count both kinds of errors.
k8-dc-refill-from-l2 [,mask= qualifier
]
Count data cache refills from L2 cache.
This event may be further qualified using
qualifier
which is a
`+'
separated set of the following keywords:
exclusive
Count operations for lines in the
``exclusive''
state.
invalid
Count operations for lines in the
``invalid''
state.
modified
Count operations for lines in the
``modified''
state.
owner
Count operations for lines in the
``owner''
state.
shared
Count operations for lines in the
``shared''
state.
The default is to count operations for lines in all the
above states.
k8-dc-refill-from-system [,mask= qualifier
]
Count data cache refills from system memory.
This event may be further qualified using
qualifier
which is a
`+'
separated set of the following keywords:
exclusive
Count operations for lines in the
``exclusive''
state.
invalid
Count operations for lines in the
``invalid''
state.
modified
Count operations for lines in the
``modified''
state.
owner
Count operations for lines in the
``owner''
state.
shared
Count operations for lines in the
``shared''
state.
The default is to count operations for lines in all the
above states.
k8-fp-dispatched-fpu-ops [,mask= 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
qualifier
which is a
`+'
separated set of the following keywords:
add-pipe-excluding-junk-ops
Count add pipe ops excluding junk ops.
add-pipe-junk-ops
Count junk ops in the add pipe.
multiply-pipe-excluding-junk-ops
Count multiply pipe ops excluding junk ops.
multiply-pipe-junk-ops
Count junk ops in the multiply pipe.
store-pipe-excluding-junk-ops
Count store pipe ops excluding junk ops
store-pipe-junk-ops
Count junk ops in the store pipe.
The default is to count all types of ops.
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.
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.
k8-fr-decoder-empty
Count cycles when there was nothing to dispatch (i.e., the decoder
was empty).
k8-fr-dispatch-stalls
Count all dispatch stalls.
k8-fr-dispatch-stall-for-segment-load
Count dispatch stalls for segment loads.
k8-fr-dispatch-stall-for-serialization
Count dispatch stalls for serialization.
k8-fr-dispatch-stall-from-branch-abort-to-retire
Count dispatch stalls from branch abort to retiral.
k8-fr-dispatch-stall-when-fpu-is-full
Count dispatch stalls when the FPU is full.
k8-fr-dispatch-stall-when-ls-is-full
Count dispatch stalls when the load/store unit is full.
k8-fr-dispatch-stall-when-reorder-buffer-is-full
Count dispatch stalls when the reorder buffer is full.
Count dispatch stalls when a far control transfer or a resync branch
is pending.
k8-fr-fpu-exceptions [,mask= qualifier
]
Count FPU exceptions.
This event is supported in revision B and later CPUs.
This event may be further qualified using
qualifier
which is a
`+'
separated set of the following keywords:
sse-and-x87-microtraps
Count SSE and x87 microtraps.
sse-reclass-microfaults
Count SSE reclass microfaults
sse-retype-microfaults
Count SSE retype microfaults
x87-reclass-microfaults
Count x87 reclass microfaults.
The default is to count all types of exceptions.
k8-fr-interrupts-masked-cycles
Count cycles when interrupts were masked (by CPU RFLAGS field IF was zero).
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).
k8-fr-number-of-breakpoints-for-dr0
Count the number of breakpoints for DR0.
k8-fr-number-of-breakpoints-for-dr1
Count the number of breakpoints for DR1.
k8-fr-number-of-breakpoints-for-dr2
Count the number of breakpoints for DR2.
k8-fr-number-of-breakpoints-for-dr3
Count the number of breakpoints for DR3.
k8-fr-retired-branches
Count retired branches including exceptions and interrupts.
k8-fr-retired-branches-mispredicted
Count mispredicted retired branches.
k8-fr-retired-far-control-transfers
Count retired far control transfers (which are always mispredicted).
Count retired fastpath double op instructions.
This event is supported in revision B and later CPUs.
This event may be further qualified using
qualifier
which is a
`+'
separated set of the following keywords:
low-op-pos-0
Count instructions with the low op in position 0.
low-op-pos-1
Count instructions with the low op in position 1.
low-op-pos-2
Count instructions with the low op in position 2.
The default is to count all types of instructions.
Count retired FPU instructions.
This event is supported in revision B and later CPUs.
This event may be further qualified using
qualifier
which is a
`+'
separated set of the following keywords:
mmx-3dnow
Count MMX and 3DNow! instructions.
packed-sse-sse2
Count packed SSE and SSE2 instructions.
scalar-sse-sse2
Count scalar SSE and SSE2 instructions
x87
Count x87 instructions.
The default is to count all types of instructions.
k8-fr-retired-near-returns
Count retired near returns.
k8-fr-retired-near-returns-mispredicted
Count mispredicted near returns.
k8-fr-retired-resyncs
Count retired resyncs (non-control transfer branches).
k8-fr-retired-taken-hardware-interrupts
Count retired taken hardware interrupts.
k8-fr-retired-taken-branches
Count retired taken branches.
k8-fr-retired-taken-branches-mispredicted
Count retired taken branches that were mispredicted.
Count memory controller bypass counter saturation events.
This event may be further qualified using
qualifier
which is a
`+'
separated set of the following keywords:
dram-controller-interface-bypass
Count DRAM controller interface bypass.
dram-controller-queue-bypass
Count DRAM controller queue bypass.
memory-controller-hi-pri-bypass
Count memory controller high priority bypasses.
memory-controller-lo-pri-bypass
Count memory controller low priority bypasses.
k8-nb-memory-controller-dram-slots-missed
Count memory controller DRAM command slots missed (in MemClks).
Count memory controller page access events.
This event may be further qualified using
qualifier
which is a
`+'
separated set of the following keywords:
page-conflict
Count page conflicts.
page-hit
Count page hits.
page-miss
Count page misses.
The default is to count all types of events.
k8-nb-memory-controller-page-table-overflow
Count memory control page table overflow events.
k8-nb-probe-result [,mask= qualifier
]
Count probe events.
This event may be further qualified using
qualifier
which is a
`+'
separated set of the following keywords:
probe-hit
Count all probe hits.
probe-hit-dirty-no-memory-cancel
Count probe hits without memory cancels.
probe-hit-dirty-with-memory-cancel
Count probe hits with memory cancels.
probe-miss
Count probe misses.
k8-nb-sized-commands [,mask= qualifier
]
Count sized commands issued.
This event may be further qualified using
qualifier
which is a
`+'
separated set of the following keywords:
Count memory control turnaround events.
This event may be further qualified using
qualifier
which is a
`+'
separated set of the following keywords:
dimm-turnaround
Count DIMM turnarounds.
read-to-write-turnaround
Count read to write turnarounds.
write-to-read-turnaround
Count write to read turnarounds.
The default is to count all types of events.
k8-nb-ht-bus0-bandwidth [,mask= qualifier
]
k8-nb-ht-bus1-bandwidth [,mask= qualifier
]
k8-nb-ht-bus2-bandwidth [,mask= qualifier
]
Count events on the HyperTransport(tm) buses.
These events may be further qualified using
qualifier
which is a
`+'
separated set of the following keywords:
buffer-release
Count buffer release messages sent.
command
Count command messages sent.
data
Count data messages sent.
nop
Count nop messages sent.
The default is to count all types of messages.
Intel P6 PMCS
Intel P6 PMCs are present in Intel
Pentium ProPentium IICeleronPentium III
and
Pentium M
processors.
These CPUs have two counters.
Some events may only be used on specific counters and some events are
defined only on specific processor models.
These PMCs are documented in
"IA-32 Intel(R) Architecture Software Developer's Manual""Volume 3: System Programming Guide"
"Order Number 245472-012"
2003
"Intel Corporation"
Some of these events are affected by processor errata described in
Event specifiers for Intel P6 PMCs can have the following common
qualifiers:
cmask= value
Configure the PMC to increment only if the number of configured
events measured in a cycle is greater than or equal to
value
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.
inv
Invert the sense of comparision when the
``cmask
''
qualifier is present, making the counter increment when the number of
events per cycle is less than the value specified by the
``cmask
''
qualifier.
os
Configure the PMC to count events happening at processor privilege
level 0.
umask= value
This qualifier is used to further qualify the event selected (see
below).
usr
Configure the PMC to count events occurring at privilege levels 1, 2
or 3.
If neither of the
``os
''
or
``usr
''
qualifiers are specified, the default is to enable both.
The event specifiers supported by Intel P6 PMCs are:
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.
p6-br-bac-missp-exec
(Pentium M
)
Count the number of branch instructions executed that where
mispredicted at the Front End (BAC).
p6-br-bogus
Count the number of bogus branches.
p6-br-call-exec
(Pentium M
)
Count the number of call instructions executed.
p6-br-call-missp-exec
(Pentium M
)
Count the number of call instructions executed that were mispredicted.
p6-br-cnd-exec
(Pentium M
)
Count the number of conditional branch instructions executed.
p6-br-cnd-missp-exec
(Pentium M
)
Count the number of conditional branch instructions executed that were
mispredicted.
p6-br-ind-call-exec
(Pentium M
)
Count the number of indirect call instructions executed.
p6-br-ind-exec
(Pentium M
)
Count the number of indirect branch instructions executed.
p6-br-ind-missp-exec
(Pentium M
)
Count the number of indirect branch instructions executed that were
mispredicted.
p6-br-inst-decoded
Count the number of branch instructions decoded.
p6-br-inst-exec
(Pentium M
)
Count the number of branch instructions executed but necessarily retired.
p6-br-inst-retired
Count the number of branch instructions retired.
p6-br-miss-pred-retired
Count the number of mispredicted branch instructions retired.
p6-br-miss-pred-taken-ret
Count the number of taken mispredicted branches retired.
p6-br-missp-exec
(Pentium M
)
Count the number of branch instructions executed that were
mispredicted at execution.
p6-br-ret-bac-missp-exec
(Pentium M
)
Count the number of return instructions executed that were
mispredicted at the Front End (BAC).
p6-br-ret-exec
(Pentium M
)
Count the number of return instructions executed.
p6-br-ret-missp-exec
(Pentium M
)
Count the number of return instructions executed that were
mispredicted at execution.
p6-br-taken-retired
Count the number of taken branches retired.
p6-btb-misses
Count the number of branches for which the BTB did not produce a
prediction.
p6-bus-bnr-drv
Count the number of bus clock cycles during which this processor is
driving the BNR# pin.
p6-bus-data-rcv
Count the number of bus clock cycles during which this processor is
receiving data.
p6-bus-drdy-clocks [,umask= qualifier
]
Count the number of clocks during which DRDY# is asserted.
An additional qualifier may be specified, and comprises one of the
following keywords:
any
Count transactions generated by any agent on the bus.
self
Count transactions generated by this processor.
The default is to count operations generated by this processor.
p6-bus-hit-drv
Count the number of bus clock cycles during which this processor is
driving the HIT# pin.
p6-bus-hitm-drv
Count the number of bus clock cycles during which this processor is
driving the HITM# pin.
p6-bus-lock-clocks [,umask= 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:
any
Count transactions generated by any agent on the bus.
self
Count transactions generated by this processor.
The default is to count operations generated by this processor.
p6-bus-req-outstanding
Count the number of bus requests outstanding in any given cycle.
p6-bus-snoop-stall
Count the number of clock cycles during which the bus is snoop stalled.
p6-bus-tran-any [,umask= qualifier
]
Count the number of completed bus transactions of any kind.
An additional qualifier may be specified and comprises one of the following
keywords:
any
Count transactions generated by any agent on the bus.
self
Count transactions generated by this processor.
The default is to count operations generated by this processor.
p6-bus-tran-brd [,umask= qualifier
]
Count the number of burst read transactions.
An additional qualifier may be specified and comprises one of the following
keywords:
any
Count transactions generated by any agent on the bus.
self
Count transactions generated by this processor.
The default is to count operations generated by this processor.
p6-bus-tran-burst [,umask= qualifier
]
Count the number of completed burst transactions.
An additional qualifier may be specified and comprises one of the following
keywords:
any
Count transactions generated by any agent on the bus.
self
Count transactions generated by this processor.
The default is to count operations generated by this processor.
p6-bus-tran-def [,umask= qualifier
]
Count the number of completed deferred transactions.
An additional qualifier may be specified and comprises one of the following
keywords:
any
Count transactions generated by any agent on the bus.
self
Count transactions generated by this processor.
The default is to count operations generated by this processor.
p6-bus-tran-ifetch [,umask= qualifier
]
Count the number of completed instruction fetch transactions.
An additional qualifier may be specified and comprises one of the following
keywords:
any
Count transactions generated by any agent on the bus.
self
Count transactions generated by this processor.
The default is to count operations generated by this processor.
p6-bus-tran-inval [,umask= qualifier
]
Count the number of completed invalidate transactions.
An additional qualifier may be specified and comprises one of the following
keywords:
any
Count transactions generated by any agent on the bus.
self
Count transactions generated by this processor.
The default is to count operations generated by this processor.
p6-bus-tran-mem [,umask= qualifier
]
Count the number of completed memory transactions.
An additional qualifier may be specified and comprises one of the following
keywords:
any
Count transactions generated by any agent on the bus.
self
Count transactions generated by this processor.
The default is to count operations generated by this processor.
p6-bus-tran-pwr [,umask= qualifier
]
Count the number of completed partial write transactions.
An additional qualifier may be specified and comprises one of the following
keywords:
any
Count transactions generated by any agent on the bus.
self
Count transactions generated by this processor.
The default is to count operations generated by this processor.
p6-bus-tran-rfo [,umask= qualifier
]
Count the number of completed read-for-ownership transactions.
An additional qualifier may be specified and comprises one of the following
keywords:
any
Count transactions generated by any agent on the bus.
self
Count transactions generated by this processor.
The default is to count operations generated by this processor.
p6-bus-trans-io [,umask= qualifier
]
Count the number of completed I/O transactions.
An additional qualifier may be specified and comprises one of the following
keywords:
any
Count transactions generated by any agent on the bus.
self
Count transactions generated by this processor.
The default is to count operations generated by this processor.
p6-bus-trans-p [,umask= qualifier
]
Count the number of completed partial transactions.
An additional qualifier may be specified and comprises one of the following
keywords:
any
Count transactions generated by any agent on the bus.
self
Count transactions generated by this processor.
The default is to count operations generated by this processor.
p6-bus-trans-wb [,umask= qualifier
]
Count the number of completed write-back transactions.
An additional qualifier may be specified and comprises one of the following
keywords:
any
Count transactions generated by any agent on the bus.
self
Count transactions generated by this processor.
The default is to count operations generated by this processor.
p6-cpu-clk-unhalted
Count the number of cycles during with the processor was not halted.
(Pentium M
)
Count the number of cycles during with the processor was not halted
and not in a thermal trip.
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.
p6-cycles-in-pending-and-masked
Count the number of processor cycles for which interrupts were
disabled and interrupts were pending.
p6-cycles-int-masked
Count the number of processor cycles for which interrupts were
disabled.
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.
p6-dcu-lines-in
Count the total lines allocated in the data cache unit.
p6-dcu-m-lines-in
Count the number of M state lines allocated in the data cache unit.
p6-dcu-m-lines-out
Count the number of M state lines evicted from the data cache unit.
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.
p6-div
Count the number of integer and floating-point divides including
speculative divides.
This event is only allocated on counter 1.
p6-emon-esp-uops
(Pentium M
)
Count the total number of micro-ops.
p6-emon-est-trans [,umask= qualifier
]
(Pentium M
)
Count the number of
Enhanced Intel SpeedStep
transitions.
An additional qualifier may be specified, and can be one of the
following keywords:
all
Count all transitions.
freq
Count only frequency transitions.
The default is to count all transitions.
p6-emon-fused-uops-ret [,umask= qualifier
]
(Pentium M
)
Count the number of retired fused micro-ops.
An additional qualifier may be specified, and may be one of the
following keywords:
all
Count all fused micro-ops.
loadop
Count only load and op micro-ops.
stdsta
Count only STD/STA micro-ops.
The default is to count all fused micro-ops.
p6-emon-kni-comp-inst-ret
(Pentium III
)
Count the number of SSE computational instructions retired.
An additional qualifier may be specified, and comprises one of the
following keywords:
packed-and-scalar
Count packed and scalar operations.
scalar
Count scalar operations only.
The default is to count packed and scalar operations.
p6-emon-kni-inst-retired [,umask= qualifier
]
(Pentium III
)
Count the number of SSE instructions retired.
An additional qualifier may be specified, and comprises one of the
following keywords:
packed-and-scalar
Count packed and scalar operations.
scalar
Count scalar operations only.
The default is to count packed and scalar operations.
p6-emon-kni-pref-dispatched [,umask= qualifier
]
(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:
nta
Count non-temporal prefetches.
t1
Count prefetches to L1.
t2
Count prefetches to L2.
wos
Count weakly ordered stores.
The default is to count non-temporal prefetches.
p6-emon-kni-pref-miss [,umask= qualifier
]
(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:
nta
Count non-temporal prefetches.
t1
Count prefetches to L1.
t2
Count prefetches to L2.
wos
Count weakly ordered stores.
The default is to count non-temporal prefetches.
p6-emon-pref-rqsts-dn
(Pentium M
)
Count the number of downward prefetches issued.
p6-emon-pref-rqsts-up
(Pentium M
)
Count the number of upward prefetches issued.
p6-emon-simd-instr-retired
(Pentium M
)
Count the number of retired
MMX
instructions.
(Pentium M
)
Count the number of computational SSE instructions retired.
An additional qualifier may be specified and can be one of the
following keywords:
sse-packed-single
Count SSE packed-single instructions.
sse-scalar-single
Count SSE scalar-single instructions.
sse2-packed-double
Count SSE2 packed-double instructions.
sse2-scalar-double
Count SSE2 scalar-double instructions.
The default is to count SSE packed-single instructions.
p6-emon-sse-sse2-inst-retired [,umask= qualifer
]
(Pentium M
)
Count the number of SSE instructions retired.
An additional qualifier can be specified, and can be one of the
following keywords:
sse-packed-single
Count SSE packed-single instructions.
sse-packed-single-scalar-single
Count SSE packed-single and scalar-single instructions.
sse2-packed-double
Count SSE2 packed-double instructions.
sse2-scalar-double
Count SSE2 scalar-double instructions.
The default is to count SSE packed-single instructions.
p6-emon-synch-uops
(Pentium M
)
Count the number of sync micro-ops.
p6-emon-thermal-trip
(Pentium M
)
Count the duration or occurrences of thermal trips.
Use the
``edge
''
qualifier to count occurrences of thermal trips.
p6-emon-unfusion
(Pentium M
)
Count the number of unfusion events in the reorder buffer.
p6-flops
Count the number of computational floating point operations retired.
This event is only allocated on counter 0.
p6-fp-assist
Count the number of floating point exceptions handled by microcode.
This event is only allocated on counter 1.
p6-fp-comps-ops-exe
Count the number of computation floating point operations executed.
This event is only allocated on counter 0.
p6-fp-mmx-trans [,umask= qualifier
]
(Pentium II , 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:
mmxtofp
Count transitions from MMX instructions to floating-point instructions.
fptommx
Count transitions from floating-point instructions to MMX instructions.
The default is to count MMX to floating-point transitions.
p6-hw-int-rx
Count the number of hardware interrupts received.
p6-ifu-fetch
Count the number of instruction fetches, both cacheable and non-cacheable.
p6-ifu-fetch-miss
Count the number of instruction fetch misses (i.e., those that produce
memory accesses).
p6-ifu-mem-stall
Count the number of cycles instruction fetch is stalled for any reason.
p6-ild-stall
Count the number of cycles the instruction length decoder is stalled.
p6-inst-decoded
Count the number of instructions decoded.
p6-inst-retired
Count the number of instructions retired.
p6-itlb-miss
Count the number of instruction TLB misses.
p6-l2-ads
Count the number of L2 address strobes.
p6-l2-dbus-busy
Count the number of cycles during which the L2 cache data bus was busy.
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.
p6-l2-ifetch [,umask= qualifier
]
Count the number of L2 instruction fetches.
An additional qualifier may be specified and comprises a list of the following
keywords separated by
`+'
characters:
e
Count operations affecting E (exclusive) state lines.
i
Count operations affecting I (invalid) state lines.
m
Count operations affecting M (modified) state lines.
s
Count operations affecting S (shared) state lines.
The default is to count operations affecting all (MESI) state lines.
p6-l2-ld [,umask= qualifier
]
Count the number of L2 data loads.
An additional qualifier may be specified and comprises a list of the following
keywords separated by
`+'
characters:
both
(Pentium M
)
Count both hardware-prefetched lines and non-hardware-prefetched lines.
e
Count operations affecting E (exclusive) state lines.
hw
(Pentium M
)
Count hardware-prefetched lines only.
i
Count operations affecting I (invalid) state lines.
m
Count operations affecting M (modified) state lines.
nonhw
(Pentium M
)
Exclude hardware-prefetched lines.
s
Count operations affecting S (shared) state lines.
The default on processors other than
Pentium M
processors is to count operations affecting all (MESI) state lines.
The default on
Pentium M
processors is to count both hardware-prefetched and
non-hardware-prefetch operations on all (MESI) state lines.
(Errata)
This event is affected by processor errata E53.
p6-l2-lines-in [,umask= qualifier
]
Count the number of L2 lines allocated.
An additional qualifier may be specified and comprises a list of the following
keywords separated by
`+'
characters:
both
(Pentium M
)
Count both hardware-prefetched lines and non-hardware-prefetched lines.
e
Count operations affecting E (exclusive) state lines.
hw
(Pentium M
)
Count hardware-prefetched lines only.
i
Count operations affecting I (invalid) state lines.
m
Count operations affecting M (modified) state lines.
nonhw
(Pentium M
)
Exclude hardware-prefetched lines.
s
Count operations affecting S (shared) state lines.
The default on processors other than
Pentium M
processors is to count operations affecting all (MESI) state lines.
The default on
Pentium M
processors is to count both hardware-prefetched and
non-hardware-prefetch operations on all (MESI) state lines.
(Errata)
This event is affected by processor errata E45.
p6-l2-lines-out [,umask= qualifier
]
Count the number of L2 lines evicted.
An additional qualifier may be specified and comprises a list of the following
keywords separated by
`+'
characters:
both
(Pentium M
)
Count both hardware-prefetched lines and non-hardware-prefetched lines.
e
Count operations affecting E (exclusive) state lines.
hw
(Pentium M
)
Count hardware-prefetched lines only.
i
Count operations affecting I (invalid) state lines.
m
Count operations affecting M (modified) state lines.
nonhw
(Pentium M only
)
Exclude hardware-prefetched lines.
s
Count operations affecting S (shared) state lines.
The default on processors other than
Pentium M
processors is to count operations affecting all (MESI) state lines.
The default on
Pentium M
processors is to count both hardware-prefetched and
non-hardware-prefetch operations on all (MESI) state lines.
(Errata)
This event is affected by processor errata E45.
p6-l2-m-lines-inm
Count the number of modified lines allocated in L2 cache.
p6-l2-m-lines-outm [,umask= qualifier
]
Count the number of L2 M-state lines evicted.
(Pentium M
)
On these processors an additional qualifier may be specified and
comprises a list of the following keywords separated by
`+'
characters:
both
Count both hardware-prefetched lines and non-hardware-prefetched lines.
hw
Count hardware-prefetched lines only.
nonhw
Exclude hardware-prefetched lines.
The default is to count both hardware-prefetched and
non-hardware-prefetch operations.
(Errata)
This event is affected by processor errata E53.
p6-l2-rqsts [,umask= qualifier
]
Count the total number of L2 requests.
An additional qualifier may be specified and comprises a list of the following
keywords separated by
`+'
characters:
e
Count operations affecting E (exclusive) state lines.
i
Count operations affecting I (invalid) state lines.
m
Count operations affecting M (modified) state lines.
s
Count operations affecting S (shared) state lines.
The default is to count operations affecting all (MESI) state lines.
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
`+'
characters:
e
Count operations affecting E (exclusive) state lines.
i
Count operations affecting I (invalid) state lines.
m
Count operations affecting M (modified) state lines.
s
Count operations affecting S (shared) state lines.
The default is to count operations affecting all (MESI) state lines.
p6-ld-blocks
Count the number of load operations delayed due to store buffer blocks.
p6-misalign-mem-ref
Count the number of misaligned data memory references (crossing a 64
bit boundary).
p6-mmx-assist
(Pentium II , Pentium III
)
Count the number of MMX assists executed.
p6-mmx-instr-exec
(Celeron , Pentium II
)
Count the number of MMX instructions executed, except MOVQ and MOVD
stores from register to memory.
p6-mmx-instr-ret
(Pentium II
)
Count the number of MMX instructions retired.
p6-mmx-instr-type-exec [,umask= qualifier
]
(Pentium II , 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
`+'
characters:
pack
Count MMX pack operation instructions.
packed-arithmetic
Count MMX packed arithmetic instructions.
packed-logical
Count MMX packed logical instructions.
packed-multiply
Count MMX packed multiply instructions.
packed-shift
Count MMX packed shift instructions.
unpack
Count MMX unpack operation instructions.
The default is to count all operations.
p6-mmx-sat-instr-exec
(Pentium II , Pentium III
)
Count the number of MMX saturating instructions executed.
p6-mmx-uops-exec
(Pentium II , Pentium III
)
Count the number of MMX micro-ops executed.
p6-mul
Count the number of integer and floating-point multiplies, including
speculative multiplies.
This event is only allocated on counter 1.
p6-partial-rat-stalls
Count the number of cycles or events for partial stalls.
p6-resource-stalls
Count the number of cycles there was a resource related stall of any kind.
p6-ret-seg-renames
(Pentium II , Pentium III
)
Count the number of segment register rename events retired.
p6-sb-drains
Count the number of cycles the store buffer is draining.
p6-seg-reg-renames [,umask= qualifier
]
(Pentium II , 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
`+'
characters:
ds
Count renames for segment register DS.
es
Count renames for segment register ES.
fs
Count renames for segment register FS.
gs
Count renames for segment register GS.
The default is to count operations affecting all segment registers.
p6-seg-rename-stalls
(Pentium II , 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
`+'
characters:
ds
Count stalls for segment register DS.
es
Count stalls for segment register ES.
fs
Count stalls for segment register FS.
gs
Count stalls for segment register GS.
The default is to count operations affecting all the segment registers.
p6-segment-reg-loads
Count the number of segment register loads.
p6-uops-retired
Count the number of micro-ops retired.
Intel P4 PMCS
Intel P4 PMCs are present in Intel
Pentium 4
and
Xeon
processors.
These PMCs are documented in
"IA-32 Intel(R) Architecture Software Developer's Manual""Volume 3: System Programming Guide"
"Order Number 245472-012"
2003
"Intel Corporation"
Further information about using these PMCs may be found in
Event specifiers for Intel P4 PMCs can have the following common
qualifiers:
active= choice
(On P4 HTT CPUs) Filter event counting based on which logical
processors are active.
The allowed values of
choice
are:
any
Count when either logical processor is active.
both
Count when both logical processors are active.
none
Count only when neither logical processor is active.
single
Count only when one logical processor is active.
The default is
``both
''
cascade
Configure the PMC to cascade onto its partner.
See
Sx Cascading P4 PMCs
below for more information.
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.
complement
Configure the counter to increment only when the event count seen is
less than the threshold qualifier value specified.
mask= 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.
os
Configure the PMC to count when the CPL of the processor is 0.
precise
Select precise event based sampling.
Precise sampling is supported by the hardware for a limited set of
events.
tag= value
Configure the PMC to tag the internal uop selected by the other
fields in this event specifier with value
value
This feature is used when cascading PMCs.
threshold= value
Configure the PMC to increment only when the event counts seen are
greater than the specified threshold value
value
usr
Configure the PMC to count when the CPL of the processor is 1, 2 or 3.
If neither of the
``os
''
or
``usr
''
qualifiers are specified, the default is to enable both.
On Intel Pentium 4 processors with HTT, events are
divided into two classes:
"TS Events"
are those where hardware can differentiate between events
generated on one logical processor from those generated on the
other.
"TI Events"
are those where hardware cannot differentiate between events
generated by multiple logical processors in a package.
Only TS events are allowed for use with process-mode PMCs on
Pentium-4/HTT CPUs.
The event specifiers supported by Intel P4 PMCs are:
p4-128bit-mmx-uop [,mask= flags
]
(TI event)
Count integer SIMD SSE2 instructions that operate on 128 bit SIMD
operands.
Qualifier
flags
can take the following value (which is also the default):
all
Count all uops operating on 128 bit SIMD integer operands in memory or
XMM register.
If an instruction contains more than one 128 bit MMX uop, then each
uop will be counted.
p4-64bit-mmx-uop [,mask= flags
]
(TI event)
Count MMX instructions that operate on 64 bit SIMD operands.
Qualifier
flags
can take the following value (which is also the default):
all
Count all uops operating on 64 bit SIMD integer operands in memory or
in MMX registers.
If an instruction contains more than one 64 bit MMX uop, then each
uop will be counted.
p4-b2b-cycles
(TI event)
Count back-to-back bys cycles.
Further documentation for this event is unavailable.
p4-bnr
(TI event)
Count bus-not-ready conditions.
Further documentation for this event is unavailable.
p4-bpu-fetch-request [,mask= qualifier
]
(TS event)
Count instruction fetch requests qualified by additional
flags specified in
qualifier
At this point only one flag is supported:
tcmiss
Count trace cache lookup misses.
The default qualifier is also
``mask=tcmiss
''
p4-branch-retired [,mask= flags
]
(TS event)
Counts retired branches.
Qualifier
flags
is a list of the following
`+'
separated strings:
mmnp
Count branches not-taken and predicted.
mmnm
Count branches not-taken and mis-predicted.
mmtp
Count branches taken and predicted.
mmtm
Count branches taken and mis-predicted.
The default qualifier counts all four kinds of branches.
p4-bsq-active-entries [,mask= qualifier
]
(TS event)
Count the number of entries (clipped at 15) currently active in the
BSQ.
Qualifier
qualifier
is a
`+'
separated set of the following flags:
req-type0 , req-type1
Forms a 2-bit number used to select the request type encoding:
0
reads excluding read invalidate
1
read invalidates
2
writes other than writebacks
3
writebacks
Bit
``req-type1
''
is the MSB for this two bit number.
req-len0 , req-len1
Forms a two-bit number that specifies the request length encoding:
0
0 chunks
1
1 chunk
3
8 chunks
Bit
``req-len1
''
is the MSB for this two bit number.
req-io-type
Count requests that are input or output requests.
req-lock-type
Count requests that lock the bus.
req-lock-cache
Count requests that lock the cache.
req-split-type
Count requests that is a bus 8-byte chunk that is split across an
8-byte boundary.
req-dem-type
Count requests that are demand (not prefetches) if set.
Count requests that are prefetches if not set.
req-ord-type
Count requests that are ordered.
mem-type0 , mem-type1 , mem-type2
Forms a 3-bit number that specifies a memory type encoding:
0
UC
1
USWC
4
WT
5
WP
6
WB
Bit
``mem-type2
''
is the MSB of this 3-bit number.
The default qualifier has all the above bits set.
Edge triggering using the
``edge
''
qualifier should not be used with this event when counting cycles.
p4-bsq-allocation [,mask= qualifier
]
(TS event)
Count allocations in the bus sequence unit according to the flags
specified in
qualifier
which is a
`+'
separated set of the following flags:
req-type0 , req-type1
Forms a 2-bit number used to select the request type encoding:
0
reads excluding read invalidate
1
read invalidates
2
writes other than writebacks
3
writebacks
Bit
``req-type1
''
is the MSB for this two bit number.
req-len0 , req-len1
Forms a two-bit number that specifies the request length encoding:
0
0 chunks
1
1 chunk
3
8 chunks
Bit
``req-len1
''
is the MSB for this two bit number.
req-io-type
Count requests that are input or output requests.
req-lock-type
Count requests that lock the bus.
req-lock-cache
Count requests that lock the cache.
req-split-type
Count requests that is a bus 8-byte chunk that is split across an
8-byte boundary.
req-dem-type
Count requests that are demand (not prefetches) if set.
Count requests that are prefetches if not set.
req-ord-type
Count requests that are ordered.
mem-type0 , mem-type1 , mem-type2
Forms a 3-bit number that specifies a memory type encoding:
0
UC
1
USWC
4
WT
5
WP
6
WB
Bit
``mem-type2
''
is the MSB of this 3-bit number.
The default qualifier has all the above bits set.
This event is usually used along with the
``edge
''
qualifier to avoid multiple counting.
p4-bsq-cache-reference [,mask= qualifier
]
(TS event)
Count cache references as seen by the bus unit (2nd or 3rd level
cache references).
Qualifier
qualifier
is a
`+'
separated list of the following keywords:
rd-2ndl-hits
Count 2nd level cache hits in the shared state.
rd-2ndl-hite
Count 2nd level cache hits in the exclusive state.
rd-2ndl-hitm
Count 2nd level cache hits in the modified state.
rd-3rdl-hits
Count 3rd level cache hits in the shared state.
rd-3rdl-hite
Count 3rd level cache hits in the exclusive state.
rd-3rdl-hitm
Count 3rd level cache hits in the modified state.
rd-2ndl-miss
Count 2nd level cache misses.
rd-3rdl-miss
Count 3rd level cache misses.
wr-2ndl-miss
Count write-back lookups from the data access cache that miss the 2nd
level cache.
The default is to count all the above events.
p4-execution-event [,mask= flags
]
(TS event)
Count the retirement of tagged uops selected through the execution
tagging mechanism.
Qualifier
flags
can contain the following strings separated by
`+'
characters:
nbogus0 , nbogus1 , nbogus2 , nbogus3
The marked uops are not bogus.
bogus0 , bogus1 , bogus2 , bogus3
The marked uops are bogus.
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.
p4-front-end-event [,mask= flags
]
(TS event)
Count the retirement of tagged uops selected through the front-end
tagging mechanism.
Qualifier
flags
can contain the following strings separated by
`+'
characters:
nbogus
The marked uops are not bogus.
bogus
The marked uops are bogus.
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.
p4-fsb-data-activity [,mask= flags
]
(TI event)
Count each DBSY or DRDY event selected by qualifier
flags
Qualifier
flags
is a
`+'
separated set of the following flags:
drdy-drv
Count when this processor is driving data onto the bus.
drdy-own
Count when this processor is reading data from the bus.
drdy-other
Count when data is on the bus but not being sampled by this processor.
dbsy-drv
Count when this processor reserves the bus for use in the next cycle
in order to drive data.
dbsy-own
Count when some agent reserves the bus for use in the next bus cycle
to drive data that this processor will sample.
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.
Flags
``drdy-own
''
and
``drdy-other
''
are mutually exclusive.
Flags
``dbsy-own
''
and
``dbsy-other
''
are mutually exclusive.
The default value for
qualifier
is
``drdy-drv+drdy-own+dbsy-drv+dbsy-own
''
p4-global-power-events [,mask= flags
]
(TS event)
Count cycles during which the processor is not stopped.
Qualifier
flags
can take the following value (which is also the default):
running
Count cycles when the processor is active.
p4-instr-retired [,mask= flags
]
(TS event)
Count instructions retired during a clock cycle.
Qualifer
flags
comprises of the following strings separated by
`+'
characters:
nbogusntag
Count non-bogus instructions that are not tagged.
nbogustag
Count non-bogus instructions that are tagged.
bogusntag
Count bogus instructions that are not tagged.
bogustag
Count bogus instructions that are tagged.
The default qualifier counts all the above kinds of instructions.
Count various types of transactions on the bus matching the flags set
in
qualifier
and
req-type
Qualifier
qualifier
is a
`+'
separated set of the following flags:
all-read
Count read entries.
all-write
Count write entries.
mem-uc
Count entries accessing uncacheable memory.
mem-wc
Count entries accessing write-combining memory.
mem-wt
Count entries accessing write-through memory.
mem-wp
Count entries accessing write-protected memory
mem-wb
Count entries accessing write-back memory.
own
Count store requests driven by the processor (i.e., not by other
processors or by DMA).
other
Count store requests driven by other processors or by DMA.
prefetch
Include hardware and software prefetch requests in the count.
The default value for
qualifier
is to enable all the above flags.
The
req-type
qualifier is a 5-bit number can be additionally used to select a
specific bus request type.
The default is 0.
The
``edge
''
qualifier is normally used with this event to prevent multiple
counting.
The exact behaviour of this event depends on the processor revision.
p4-itlb-reference [mask= qualifier
]
(TS event)
Count translations using the intruction translation look-aside
buffer.
The
qualifier
argument is a list of the following strings separated by
`+'
characters.
hit
Count ITLB hits.
miss
Count ITLB misses.
hit-uc
Count uncacheable ITLB hits.
If no
qualifier
is specified the default is to count all the three kinds of ITLB
translations.
p4-load-port-replay [,mask= qualifier
]
(TS event)
Count replayed events at the load port.
Qualifier
qualifier
can take on one value:
split-ld
Count split loads.
The default value for
qualifier
is
``split-ld
''
p4-mispred-branch-retired [,mask= flags
]
(TS event)
Count mispredicted IA-32 branch instructions.
Qualifier
flags
can take the following value (which is also the default):
nbogus
Count non-bogus retired branch instructions.
p4-machine-clear [,mask= flags
]
(TS event)
Count the number of pipeline clears seen by the processor.
Qualifer
flags
is a list of the following strings separated by
`+'
characters:
clear
Count for a portion of the many cycles when the machine is being
cleared for any reason.
moclear
Count machine clears due to memory ordering issues.
smclear
Count machine clears due to self-modifying code.
Use qualifier
``edge
''
to get a count of occurrences of machine clears.
The default qualifier is
``clear
''
p4-memory-cancel [,mask= event-list
]
(TS event)
Count the cancelling of various kinds of requests in the data cache
address control unit of the CPU.
The qualifier
event-list
is a list of the following strings separated by
`+'
characters:
st-rb-full
Requests cancelled because no store request buffer was available.
64k-conf
Requests that conflict due to 64K aliasing.
If
event-list
is not specified, then the default is to count both kinds of events.
p4-memory-complete [,mask= event-list
]
(TS event)
Count the completion of load split, store split, uncacheable split and
uncacheable load operations selected by qualifier
event-list
The qualifier
event-list
is a
`+'
separated list of the following flags:
lsc
Count load splits completed, excluding loads from uncacheable or
write-combining areas.
ssc
Count any split stores completed.
The default is to count both kinds of operations.
p4-mob-load-replay [,mask= qualifier
]
(TS event)
Count load replays triggered by the memory order buffer.
Qualifier
qualifier
can be a
`+'
separated list of the following flags:
no-sta
Count replays because of unknown store addresses.
no-std
Count replays because of unknown store data.
partial-data
Count replays because of partially overlapped data accesses between
load and store operations.
unalgn-addr
Count replays because of mismatches in the lower 4 bits of load and
store operations.
The default qualifier is
no-sta+no-std+partial-data+unalgn-addr
p4-packed-dp-uop [,mask= flags
]
(TI event)
Count packed double-precision uops.
Qualifier
flags
can take the following value (which is also the default):
all
Count all uops operating on packed double-precision operands.
p4-packed-sp-uop [,mask= flags
]
(TI event)
Count packed single-precision uops.
Qualifier
flags
can take the following value (which is also the default):
all
Count all uops operating on packed single-precision operands.
p4-page-walk-type [,mask= qualifier
]
(TI event)
Count page walks performed by the page miss handler.
Qualifier
qualifier
can be a
`+'
separated list of the following keywords:
dtmiss
Count page walks for data TLB misses.
itmiss
Count page walks for instruction TLB misses.
The default value for
qualifier
is
``dtmiss+itmiss
''
p4-replay-event [,mask= flags
]
(TS event)
Count the retirement of tagged uops selected through the replay
tagging mechanism.
Qualifier
flags
contains a
`+'
separated set of the following strings:
nbogus
The marked uops are not bogus.
bogus
The marked uops are bogus.
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.
p4-resource-stall [,mask= flags
]
(TS event)
Count the occurrence or latency of stalls in the allocator.
Qualifier
flags
can take the following value (which is also the default):
sbfull
A stall due to the lack of store buffers.
p4-response
(TI event)
Count different types of responses.
Further documentation on this event is not available.
p4-retired-branch-type [,mask= flags
]
(TS event)
Count branches retired.
Qualifier
flags
contains a
`+'
separated list of strings:
conditional
Count conditional jumps.
call
Count direct and indirect call branches.
return
Count return branches.
indirect
Count returns, indirect calls or indirect jumps.
The default qualifier counts all the above branch types.
p4-retired-mispred-branch-type [,mask= flags
]
(TS event)
Count mispredicted branches retired.
Qualifier
flags
contains a
`+'
separated list of strings:
conditional
Count conditional jumps.
call
Count indirect call branches.
return
Count return branches.
indirect
Count returns, indirect calls or indirect jumps.
The default qualifier counts all the above branch types.
p4-scalar-dp-uop [,mask= flags
]
(TI event)
Count the number of scalar double-precision uops.
Qualifier
flags
can take the following value (which is also the default):
all
Count the number of scalar double-precision uops.
p4-scalar-sp-uop [,mask= flags
]
(TI event)
Count the number of scalar single-precision uops.
Qualifier
flags
can take the following value (which is also the default):
all
Count all uops operating on scalar single-precision operands.
p4-snoop
(TI event)
Count snoop traffic.
Further documentation on this event is not available.
p4-sse-input-assist [,mask= flags
]
(TI event)
Count the number of times an assist is required to handle problems
with the operands for SSE and SSE2 operations.
Qualifier
flags
can take the following value (which is also the default):
all
Count assists for all SSE and SSE2 uops.
p4-store-port-replay [,mask= qualifier
]
(TS event)
Count events replayed at the store port.
Qualifier
qualifier
can take on one value:
split-st
Count split stores.
The default value for
qualifier
is
``split-st
''
p4-tc-deliver-mode [,mask= qualifier
]
(TI event)
Count the duration in cycles of operating modes of the trace cache and
decode engine.
The desired operating mode is selected by
qualifier
which is a list of the following strings separated by
`+'
characters:
DD
Both logical processors are in deliver mode.
DB
Logical processor 0 is in deliver mode while logical processor 1 is in
build mode.
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.
BD
Logical processor 0 is in build mode while logical processor 1 is in
deliver mode.
BB
Both logical processors are in build mode.
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.
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.
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.
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
``DD+DB+DI+BD+BB+BI+ID+IB
''
p4-tc-ms-xfer [,mask= flags
]
(TI event)
Count the number of times uop delivery changed from the trace cache to
MS ROM.
Qualifier
flags
can take the following value (which is also the default):
cisc
Count TC to MS transfers.
p4-uop-queue-writes [,mask= flags
]
(TS event)
Count the number of valid uops written to the uop queue.
Qualifier
flags
is a list of the following strings, separated by
`+'
characters:
from-tc-build
Count uops being written from the trace cache in build mode.
from-tc-deliver
Count uops being written from the trace cache in deliver mode.
from-rom
Count uops being written from microcode ROM.
The default qualifier counts all the above kinds of uops.
p4-uop-type [,mask= flags
]
(TS event)
This event is used in conjunction with the front-end at-retirement
mechanism to tag load and store uops.
Qualifer
flags
comprises the following strings separated by
`+'
characters:
tagloads
Mark uops that are load operations.
tagstores
Mark uops that are store operations.
The default qualifier counts both kinds of uops.
p4-uops-retired [,mask= flags
]
(TS event)
Count uops retired during a clock cycle.
Qualifier
flags
comprises the following strings separated by
`+'
characters:
nbogus
Count marked uops that are not bogus.
bogus
Count marked uops that are bogus.
The default qualifier counts both kinds of uops.
p4-wc-buffer [,mask= flags
]
(TI event)
Count write-combining buffer operations.
Qualifier
flags
contains the following strings separated by
`+'
characters:
wcb-evicts
WC buffer evictions due to any cause.
wcb-full-evict
WC buffer evictions due to no WC buffer being available.
The default qualifer counts both kinds of evictions.
p4-x87-assist [,mask= flags
]
(TS event)
Count the retirement of x87 instructions that required special
handling.
Qualifier
flags
contains the following strings separated by
`+'
characters:
fpsu
Count instructions that saw an FP stack underflow.
fpso
Count instructions that saw an FP stack overflow.
poao
Count instructions that saw an x87 output overflow.
poau
Count instructions that saw an x87 output underflow.
prea
Count instructions that needed an x87 input assist.
The default qualifier counts all the above types of instruction
retirements.
p4-x87-fp-uop [,mask= flags
]
(TI event)
Count x87 floating-point uops.
Qualifier
flags
can take the following value (which is also the default):
all
Count all x87 floating-point uops.
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.
p4-x87-simd-moves-uop [,mask= flags
]
(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
flags
may contain the following keywords separated by
`+'
characters:
allp0
Count all x87 and SIMD store and move uops.
allp2
Count all x87 and SIMD load uops.
The default is to count all uops.
(Errata)
This event may be affected by processor errata N43.
Cascading P4 PMCs
PMC cascading support is currently poorly implemented.
While individual event counters may be allocated with a
``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.
Precise Event Based Sampling
Support for precise event based sampling is currently
unimplemented in
hwpmc(4).
IMPLEMENTATION NOTES
On the i386 architecture,
Fx has historically allowed the use of the RDTSC instruction from
user-mode (i.e., at a processor CPL of 3) by any process.
This behaviour is preserved by
hwpmc(4).
RETURN VALUES
The
pmc_name_of_capability (,);
pmc_name_of_class (,);
pmc_name_of_cputype (,);
pmc_name_of_disposition (,);
pmc_name_of_event (,);
pmc_name_of_mode (,);
and
pmc_name_of_state ();
functions return a pointer to the human readable form of their argument.
These pointers may point to statically allocated storage and must
not be passed to
free (.);
In case of an error, these functions return
NULL
and set the global variable
errno
The functions
pmc_ncpu ();
and
pmc_npmc ();
return the number of CPUs and number of PMCs configured respectively;
in case of an error they return the value
-1
and set the global variable
errno
All other functions return the value
0
if successful; otherwise the value
-1
is returned and the global variable
errno
is set to indicate the error.
COMPATIBILITY
The interface between the
pmc
library and the
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
hwpmc(4)
driver, applications are urged to dynamically link with the
pmc
library.
The
pmc
API is
Ud
ERRORS
A call to
pmc_init ();
may fail with the following errors in addition to those returned by
modfind(2),
modstat(2)
and
hwpmc(4):
Bq Er ENXIO
An unknown CPU type was encountered during initialization.
Bq Er EPROGMISMATCH
The version number of the
hwpmc(4)
kernel module did not match that compiled into the
pmc
library.
A call to
pmc_capabilities (,);
pmc_name_of_capability (,);
pmc_name_of_disposition (,);
pmc_name_of_state (,);
pmc_name_of_event (,);
pmc_name_of_mode ();
pmc_name_of_class ();
and
pmc_width ();
may fail with the following error:
Bq Er EINVAL
An invalid argument was passed to the function.
A call to
pmc_cpuinfo ();
or
pmc_ncpu ();
may fail with the following error:
Bq Er ENXIO
The
pmc
has not been initialized.
A call to
pmc_npmc ();
may fail with the following errors:
Bq Er EINVAL
The argument passed in was out of range.
Bq Er ENXIO
The
pmc
library has not been initialized.
A call to
pmc_pmcinfo ();
may fail with the following errors, in addition to those returned by
hwpmc(4):
Bq Er ENXIO
The
pmc
library is not yet initialized.
A call to
pmc_allocate ();
may fail with the following errors, in addition to those returned by
hwpmc(4):
Bq Er EINVAL
The
Fa mode
argument passed in had an illegal value, or the event specification
Fa ctrspec
was unrecognized for this CPU type.
Calls to
pmc_attach (,);
pmc_configure_logfile (,);
pmc_detach (,);
pmc_disable (,);
pmc_enable (,);
pmc_get_driver_stats (,);
pmc_get_msr (,);
pmc_read (,);
pmc_release (,);
pmc_rw (,);
pmc_set (,);
pmc_start (,);
pmc_stop (,);
pmc_write (,);
and
pmc_writelog ();
may fail with the errors described in
hwpmc(4).
If a log file was configured using
pmc_configure_logfile ();
and the
hwpmc(4)
driver encountered an error while logging data to it, then
logging will be stopped and a subsequent call to
pmc_flush_logfile ();
will fail with the error code seen by the
hwpmc(4)
driver.
The information returned by
pmc_cpuinfo (,);
pmc_ncpu ();
and possibly
pmc_npmc ();
should really be available all the time, through a better designed
interface and not just when
hwpmc(4)
is present in the kernel.
