5edfb77dd3
New kernel events can be added at various location for sampling or counting. This will for example allow easy system profiling whatever the processor is with known tools like pmcstat(8). Simultaneous usage of software PMC and hardware PMC is possible, for example looking at the lock acquire failure, page fault while sampling on instructions. Sponsored by: NETASQ MFC after: 1 month
1226 lines
31 KiB
Groff
1226 lines
31 KiB
Groff
.\" 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 THE AUTHOR AND CONTRIBUTORS ``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 THE AUTHOR OR CONTRIBUTORS 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 October 4, 2008
|
|
.Dt PMC.P4 3
|
|
.Os
|
|
.Sh NAME
|
|
.Nm pmc.p4
|
|
.Nd measurement events for
|
|
.Tn "Intel Pentium 4"
|
|
and other
|
|
.Tn Netburst
|
|
architecture CPUs
|
|
.Sh LIBRARY
|
|
.Lb libpmc
|
|
.Sh SYNOPSIS
|
|
.In pmc.h
|
|
.Sh DESCRIPTION
|
|
Intel P4 PMCs are present in Intel
|
|
.Tn "Pentium 4"
|
|
and
|
|
.Tn Xeon
|
|
processors that use the
|
|
.Tn Netburst
|
|
CPU architecture.
|
|
.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
|
|
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
|
|
.Ss PMC Features
|
|
Intel Pentium 4 PMCs are 40 bits wide.
|
|
Each CPU contains 18 PMCs, divided into 4 groups with 4, 4, 4 and 6
|
|
PMCs respectively.
|
|
On processors with hyperthreading support, PMC resources are shared
|
|
between logical processors.
|
|
These PMCs support the following capabilities:
|
|
.Bl -column "PMC_CAP_INTERRUPT" "Support"
|
|
.It Em Capability Ta Em Support
|
|
.It PMC_CAP_CASCADE Ta Yes
|
|
.It PMC_CAP_EDGE Ta Yes
|
|
.It PMC_CAP_INTERRUPT Ta Yes
|
|
.It PMC_CAP_INVERT Ta Yes
|
|
.It PMC_CAP_READ Ta Yes
|
|
.It PMC_CAP_PRECISE Ta Unimplemented
|
|
.It PMC_CAP_SYSTEM Ta Yes
|
|
.It PMC_CAP_TAGGING Ta Yes
|
|
.It PMC_CAP_THRESHOLD Ta Yes
|
|
.It PMC_CAP_USER Ta Yes
|
|
.It PMC_CAP_WRITE Ta Yes
|
|
.El
|
|
.Ss Event Qualifiers
|
|
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
|
|
comparison 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:
|
|
.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.
|
|
Qualifier
|
|
.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 un-cacheable 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 behavior 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 un-cacheable 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 behavior 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 instruction 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 un-cacheable 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.
|
|
Qualifier
|
|
.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 canceling 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, un-cacheable split and
|
|
un-cacheable 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 un-cacheable 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.
|
|
Qualifier
|
|
.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 qualifier 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.
|
|
.Ss Event Name Aliases
|
|
The following table shows the mapping between the PMC-independent
|
|
aliases supported by
|
|
.Lb libpmc
|
|
and the underlying hardware events used.
|
|
.Bl -column "branch-mispredicts" "Description"
|
|
.It Em Alias Ta Em Event
|
|
.It Li branches Ta Li p4-branch-retired,mask=mmtp+mmtm
|
|
.It Li branch-mispredicts Ta Li p4-mispred-branch-retired
|
|
.It Li dc-misses Ta (unsupported)
|
|
.It Li ic-misses Ta (unsupported)
|
|
.It Li instructions Ta Li p4-instr-retired,mask=nbogusntag+nbogustag
|
|
.It Li interrupts Ta Li (unsupported)
|
|
.It Li unhalted-cycles Ta Li p4-global-power-events
|
|
.El
|
|
.Sh SEE ALSO
|
|
.Xr pmc 3 ,
|
|
.Xr pmc.atom 3 ,
|
|
.Xr pmc.core 3 ,
|
|
.Xr pmc.core2 3 ,
|
|
.Xr pmc.iaf 3 ,
|
|
.Xr pmc.k7 3 ,
|
|
.Xr pmc.k8 3 ,
|
|
.Xr pmc.p5 3 ,
|
|
.Xr pmc.p6 3 ,
|
|
.Xr pmc.soft 3 ,
|
|
.Xr pmc.tsc 3 ,
|
|
.Xr pmclog 3 ,
|
|
.Xr hwpmc 4
|
|
.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 .
|