freebsd-skq/lib/libpmc/pmc.p4.3
fabient 5edfb77dd3 Add software PMC support.
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
2012-03-28 20:58:30 +00:00

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 .