Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions
are met:
1. Redistributions of source code must retain the above copyright
notice, this list of conditions and the following disclaimer.
2. Redistributions in binary form must reproduce the above copyright
notice, this list of conditions and the following disclaimer in the
documentation and/or other materials provided with the distribution.
This software is provided by Joseph Koshy ``as is'' and
any express or implied warranties, including, but not limited to, the
implied warranties of merchantability and fitness for a particular purpose
are disclaimed. in no event shall Joseph Koshy be liable
for any direct, indirect, incidental, special, exemplary, or consequential
damages (including, but not limited to, procurement of substitute goods
or services; loss of use, data, or profits; or business interruption)
however caused and on any theory of liability, whether in contract, strict
liability, or tort (including negligence or otherwise) arising in any way
out of the use of this software, even if advised of the possibility of
such damage.
$FreeBSD: head/lib/libpmc/pmc.k8.3 183084 2008-09-16 16:34:55Z jkoshy $
.Dd September 17, 2008 .Os .Dt PMC.K8 3 .Sh NAME .Nm pmc.k8 .Nd measurement events for .Tn AMD .Tn Athlon 64 (K8 family) CPUs .Sh LIBRARY .Lb libpmc .Sh SYNOPSIS n pmc.h .Sh DESCRIPTION AMD K8 PMCs are present in the .Tn "AMD Athlon64" and .Tn "AMD Opteron" series of CPUs. They are documented in the .Rs .%B "BIOS and Kernel Developer's Guide for the AMD Athlon(tm) 64 and AMD Opteron Processors" .%N "Publication No. 26094" .%D "April 2004" .%Q "Advanced Micro Devices, Inc." .Re .Ss PMC Features AMD K8 PMCs are 48 bits wide. Each CPU contains 4 PMCs with the following capabilities: l -column "PMC_CAP_INTERRUPT" "Support" t Em Capability Ta Em Support t PMC_CAP_CASCADE Ta No t PMC_CAP_EDGE Ta Yes t PMC_CAP_INTERRUPT Ta Yes t PMC_CAP_INVERT Ta Yes t PMC_CAP_READ Ta Yes t PMC_CAP_PRECISE Ta No t PMC_CAP_SYSTEM Ta Yes t PMC_CAP_TAGGING Ta No t PMC_CAP_THRESHOLD Ta Yes t PMC_CAP_USER Ta Yes t PMC_CAP_WRITE Ta Yes .El .Ss Event Qualifiers
p Event specifiers for AMD K8 PMCs can have the following optional qualifiers: l -tag -width indent t Li count= Ns Ar value Configure the counter to increment only if the number of configured events measured in a cycle is greater than or equal to .Ar value . t Li edge Configure the counter to only count negated-to-asserted transitions of the conditions expressed by the other fields. In other words, the counter will increment only once whenever a given condition becomes true, irrespective of the number of clocks during which the condition remains true. t Li inv Invert the sense of comparision when the .Dq Li count qualifier is present, making the counter to increment when the number of events per cycle is less than the value specified by the .Dq Li count qualifier. t Li mask= Ns Ar qualifier Many event specifiers for AMD K8 PMCs need to be additionally qualified using a mask qualifier. These additional qualifiers are event-specific and are documented along with their associated event specifiers below. t Li os Configure the PMC to count events happening at privilege level 0. t Li usr Configure the PMC to count events occurring at privilege levels 1, 2 or 3. .El
p If neither of the .Dq Li os or .Dq Li usr qualifiers were specified, the default is to enable both. .Ss AMD K8 Event Specifiers The event specifiers supported on AMD K8 PMCs are: l -tag -width indent t Li k8-bu-cpu-clk-unhalted Count the number of clock cycles when the CPU is not in the HLT or STPCLK states. t Li k8-bu-fill-request-l2-miss Op Li ,mask= Ns Ar qualifier Count fill requests that missed in the L2 cache. This event may be further qualified using .Ar qualifier , which is a .Ql + separated set of the following keywords:
p l -tag -width indent -compact t Li dc-fill Count data cache fill requests. t Li ic-fill Count instruction cache fill requests. t Li tlb-reload Count TLB reloads. .El
p The default is to count all types of requests. t Li k8-bu-internal-l2-request Op Li ,mask= Ns Ar qualifier Count internally generated requests to the L2 cache. This event may be further qualified using .Ar qualifier , which is a .Ql + separated set of the following keywords:
p l -tag -width indent -compact t Li cancelled Count cancelled requests. t Li dc-fill Count data cache fill requests. t Li ic-fill Count instruction cache fill requests. t Li tag-snoop Count tag snoop requests. t Li tlb-reload Count TLB reloads. .El
p The default is to count all types of requests. t Li k8-dc-access Count data cache accesses including microcode scratchpad accesses. t Li k8-dc-copyback Op Li ,mask= Ns Ar qualifier Count data cache copyback operations. This event may be further qualified using .Ar qualifier , which is a .Ql + separated set of the following keywords:
p l -tag -width indent -compact t Li exclusive Count operations for lines in the .Dq exclusive state. t Li invalid Count operations for lines in the .Dq invalid state. t Li modified Count operations for lines in the .Dq modified state. t Li owner Count operations for lines in the .Dq owner state. t Li shared Count operations for lines in the .Dq shared state. .El
p The default is to count operations for lines in all the above states. t Li k8-dc-dcache-accesses-by-locks Op Li ,mask= Ns Ar qualifier Count data cache accesses by lock instructions. This event is only available on processors of revision C or later vintage. This event may be further qualified using .Ar qualifier , which is a .Ql + separated set of the following keywords:
p l -tag -width indent -compact t Li accesses Count data cache accesses by lock instructions. t Li misses Count data cache misses by lock instructions. .El
p The default is to count all accesses. t Li k8-dc-dispatched-prefetch-instructions Op Li ,mask= Ns Ar qualifier Count the number of dispatched prefetch instructions. This event may be further qualified using .Ar qualifier , which is a .Ql + separated set of the following keywords:
p l -tag -width indent -compact t Li load Count load operations. t Li nta Count non-temporal operations. t Li store Count store operations. .El
p The default is to count all operations. t Li k8-dc-l1-dtlb-miss-and-l2-dtlb-hit Count L1 DTLB misses that are L2 DTLB hits. t Li k8-dc-l1-dtlb-miss-and-l2-dtlb-miss Count L1 DTLB misses that are also misses in the L2 DTLB. t Li k8-dc-microarchitectural-early-cancel-of-an-access Count microarchitectural early cancels of data cache accesses. t Li k8-dc-microarchitectural-late-cancel-of-an-access Count microarchitectural late cancels of data cache accesses. t Li k8-dc-misaligned-data-reference Count misaligned data references. t Li k8-dc-miss Count data cache misses. t Li k8-dc-one-bit-ecc-error Op Li ,mask= Ns Ar qualifier Count one bit ECC errors found by the scrubber. This event may be further qualified using .Ar qualifier , which is a .Ql + separated set of the following keywords:
p l -tag -width indent -compact t Li scrubber Count scrubber detected errors. t Li piggyback Count piggyback scrubber errors. .El
p The default is to count both kinds of errors. t Li k8-dc-refill-from-l2 Op Li ,mask= Ns Ar qualifier Count data cache refills from L2 cache. This event may be further qualified using .Ar qualifier , which is a .Ql + separated set of the following keywords:
p l -tag -width indent -compact t Li exclusive Count operations for lines in the .Dq exclusive state. t Li invalid Count operations for lines in the .Dq invalid state. t Li modified Count operations for lines in the .Dq modified state. t Li owner Count operations for lines in the .Dq owner state. t Li shared Count operations for lines in the .Dq shared state. .El
p The default is to count operations for lines in all the above states. t Li k8-dc-refill-from-system Op Li ,mask= Ns Ar qualifier Count data cache refills from system memory. This event may be further qualified using .Ar qualifier , which is a .Ql + separated set of the following keywords:
p l -tag -width indent -compact t Li exclusive Count operations for lines in the .Dq exclusive state. t Li invalid Count operations for lines in the .Dq invalid state. t Li modified Count operations for lines in the .Dq modified state. t Li owner Count operations for lines in the .Dq owner state. t Li shared Count operations for lines in the .Dq shared state. .El
p The default is to count operations for lines in all the above states. t Li k8-fp-dispatched-fpu-ops Op Li ,mask= Ns Ar qualifier Count the number of dispatched FPU ops. This event is supported in revision B and later CPUs. This event may be further qualified using .Ar qualifier , which is a .Ql + separated set of the following keywords:
p l -tag -width indent -compact t Li add-pipe-excluding-junk-ops Count add pipe ops excluding junk ops. t Li add-pipe-junk-ops Count junk ops in the add pipe. t Li multiply-pipe-excluding-junk-ops Count multiply pipe ops excluding junk ops. t Li multiply-pipe-junk-ops Count junk ops in the multiply pipe. t Li store-pipe-excluding-junk-ops Count store pipe ops excluding junk ops t Li store-pipe-junk-ops Count junk ops in the store pipe. .El
p
The default is to count all types of ops.
t Li k8-fp-cycles-with-no-fpu-ops-retired Count cycles when no FPU ops were retired.
This event is supported in revision B and later CPUs.
t Li k8-fp-dispatched-fpu-fast-flag-ops Count dispatched FPU ops that use the fast flag interface.
This event is supported in revision B and later CPUs.
t Li k8-fr-decoder-empty Count cycles when there was nothing to dispatch (i.e., the decoder
was empty).
t Li k8-fr-dispatch-stalls Count all dispatch stalls.
t Li k8-fr-dispatch-stall-for-segment-load Count dispatch stalls for segment loads.
t Li k8-fr-dispatch-stall-for-serialization Count dispatch stalls for serialization.
t Li k8-fr-dispatch-stall-from-branch-abort-to-retire Count dispatch stalls from branch abort to retiral.
t Li k8-fr-dispatch-stall-when-fpu-is-full Count dispatch stalls when the FPU is full.
t Li k8-fr-dispatch-stall-when-ls-is-full Count dispatch stalls when the load/store unit is full.
t Li k8-fr-dispatch-stall-when-reorder-buffer-is-full Count dispatch stalls when the reorder buffer is full.
t Li k8-fr-dispatch-stall-when-reservation-stations-are-full Count dispatch stalls when reservation stations are full.
t Li k8-fr-dispatch-stall-when-waiting-for-all-to-be-quiet Count dispatch stalls when waiting for all to be quiet.
XXX What does "waiting for all to be quiet" mean?
t Li k8-fr-dispatch-stall-when-waiting-far-xfer-or-resync-branch-pending Count dispatch stalls when a far control transfer or a resync branch
is pending.
t Li k8-fr-fpu-exceptions Op Li ,mask= Ns Ar qualifier Count FPU exceptions.
This event is supported in revision B and later CPUs.
This event may be further qualified using
.Ar qualifier ,
which is a
.Ql +
separated set of the following keywords:
p l -tag -width indent -compact t Li sse-and-x87-microtraps Count SSE and x87 microtraps. t Li sse-reclass-microfaults Count SSE reclass microfaults t Li sse-retype-microfaults Count SSE retype microfaults t Li x87-reclass-microfaults Count x87 reclass microfaults. .El
p The default is to count all types of exceptions. t Li k8-fr-interrupts-masked-cycles Count cycles when interrupts were masked (by CPU RFLAGS field IF was zero). t Li k8-fr-interrupts-masked-while-pending-cycles Count cycles while interrupts were masked while pending (i.e., cycles when INTR was asserted while CPU RFLAGS field IF was zero). t Li k8-fr-number-of-breakpoints-for-dr0 Count the number of breakpoints for DR0. t Li k8-fr-number-of-breakpoints-for-dr1 Count the number of breakpoints for DR1. t Li k8-fr-number-of-breakpoints-for-dr2 Count the number of breakpoints for DR2. t Li k8-fr-number-of-breakpoints-for-dr3 Count the number of breakpoints for DR3. t Li k8-fr-retired-branches Count retired branches including exceptions and interrupts. t Li k8-fr-retired-branches-mispredicted Count mispredicted retired branches. t Li k8-fr-retired-far-control-transfers Count retired far control transfers (which are always mispredicted). t Li k8-fr-retired-fastpath-double-op-instructions Op Li ,mask= Ns Ar qualifier Count retired fastpath double op instructions. This event is supported in revision B and later CPUs. This event may be further qualified using .Ar qualifier , which is a .Ql + separated set of the following keywords:
p l -tag -width indent -compact t Li low-op-pos-0 Count instructions with the low op in position 0. t Li low-op-pos-1 Count instructions with the low op in position 1. t Li low-op-pos-2 Count instructions with the low op in position 2. .El
p The default is to count all types of instructions. t Li k8-fr-retired-fpu-instructions Op Li ,mask= Ns Ar qualifier Count retired FPU instructions. This event is supported in revision B and later CPUs. This event may be further qualified using .Ar qualifier , which is a .Ql + separated set of the following keywords:
p l -tag -width indent -compact t Li mmx-3dnow Count MMX and 3DNow! instructions. t Li packed-sse-sse2 Count packed SSE and SSE2 instructions. t Li scalar-sse-sse2 Count scalar SSE and SSE2 instructions t Li x87 Count x87 instructions. .El
p The default is to count all types of instructions. t Li k8-fr-retired-near-returns Count retired near returns. t Li k8-fr-retired-near-returns-mispredicted Count mispredicted near returns. t Li k8-fr-retired-resyncs Count retired resyncs (non-control transfer branches). t Li k8-fr-retired-taken-hardware-interrupts Count retired taken hardware interrupts. t Li k8-fr-retired-taken-branches Count retired taken branches. t Li k8-fr-retired-taken-branches-mispredicted Count retired taken branches that were mispredicted. t Li k8-fr-retired-taken-branches-mispredicted-by-addr-miscompare Count retired taken branches that were mispredicted only due to an address miscompare. t Li k8-fr-retired-uops Count retired uops. t Li k8-fr-retired-x86-instructions Count retired x86 instructions including exceptions and interrupts. t Li k8-ic-fetch Count instruction cache fetches. t Li k8-ic-instruction-fetch-stall Count cycles in stalls due to instruction fetch. t Li k8-ic-l1-itlb-miss-and-l2-itlb-hit Count L1 ITLB misses that are L2 ITLB hits. t Li k8-ic-l1-itlb-miss-and-l2-itlb-miss Count ITLB misses that miss in both L1 and L2 ITLBs. t Li k8-ic-microarchitectural-resync-by-snoop Count microarchitectural resyncs caused by snoops. t Li k8-ic-miss Count instruction cache misses. t Li k8-ic-refill-from-l2 Count instruction cache refills from L2 cache. t Li k8-ic-refill-from-system Count instruction cache refills from system memory. t Li k8-ic-return-stack-hits Count hits to the return stack. t Li k8-ic-return-stack-overflow Count overflows of the return stack. t Li k8-ls-buffer2-full Count load/store buffer2 full events. t Li k8-ls-locked-operation Op Li ,mask= Ns Ar qualifier Count locked operations. For revision C and later CPUs, the following qualifiers are supported:
p l -tag -width indent -compact t Li cycles-in-request Count the number of cycles in the lock request/grant stage. t Li cycles-to-complete Count the number of cycles a lock takes to complete once it is non-speculative and is the older load/store operation. t Li locked-instructions Count the number of lock instructions executed. .El
p
The default is to count the number of lock instructions executed.
t Li k8-ls-microarchitectural-late-cancel Count microarchitectural late cancels of operations in the load/store
unit.
t Li k8-ls-microarchitectural-resync-by-self-modifying-code Count microarchitectural resyncs caused by self-modifying code.
t Li k8-ls-microarchitectural-resync-by-snoop Count microarchitectural resyncs caused by snoops.
t Li k8-ls-retired-cflush-instructions Count retired CFLUSH instructions.
t Li k8-ls-retired-cpuid-instructions Count retired CPUID instructions.
t Li k8-ls-segment-register-load Op Li ,mask= Ns Ar qualifier Count segment register loads.
This event may be further qualified using
.Ar qualifier ,
which is a
.Ql +
separated set of the following keywords:
l -tag -width indent -compact t Li cs Count CS register loads.
t Li ds Count DS register loads.
t Li es Count ES register loads.
t Li fs Count FS register loads.
t Li gs Count GS register loads.
.It Li hs
Count HS register loads.
XXX "HS" register?
t Li ss Count SS register loads.
.El
p The default is to count all types of loads. t Li k8-nb-memory-controller-bypass-saturation Op Li ,mask= Ns Ar qualifier Count memory controller bypass counter saturation events. This event may be further qualified using .Ar qualifier , which is a .Ql + separated set of the following keywords:
p l -tag -width indent -compact t Li dram-controller-interface-bypass Count DRAM controller interface bypass. t Li dram-controller-queue-bypass Count DRAM controller queue bypass. t Li memory-controller-hi-pri-bypass Count memory controller high priority bypasses. t Li memory-controller-lo-pri-bypass Count memory controller low priority bypasses. .El
p t Li k8-nb-memory-controller-dram-slots-missed Count memory controller DRAM command slots missed (in MemClks). t Li k8-nb-memory-controller-page-access-event Op Li ,mask= Ns Ar qualifier Count memory controller page access events. This event may be further qualified using .Ar qualifier , which is a .Ql + separated set of the following keywords:
p l -tag -width indent -compact t Li page-conflict Count page conflicts. t Li page-hit Count page hits. t Li page-miss Count page misses. .El
p The default is to count all types of events. t Li k8-nb-memory-controller-page-table-overflow Count memory control page table overflow events. t Li k8-nb-probe-result Op Li ,mask= Ns Ar qualifier Count probe events. This event may be further qualified using .Ar qualifier , which is a .Ql + separated set of the following keywords:
p l -tag -width indent -compact t Li probe-hit Count all probe hits. t Li probe-hit-dirty-no-memory-cancel Count probe hits without memory cancels. t Li probe-hit-dirty-with-memory-cancel Count probe hits with memory cancels. t Li probe-miss Count probe misses. .El t Li k8-nb-sized-commands Op Li ,mask= Ns Ar qualifier Count sized commands issued. This event may be further qualified using .Ar qualifier , which is a .Ql + separated set of the following keywords:
p l -tag -width indent -compact t Li nonpostwrszbyte t Li nonpostwrszdword t Li postwrszbyte t Li postwrszdword t Li rdszbyte t Li rdszdword t Li rdmodwr .El
p The default is to count all types of commands. t Li k8-nb-memory-controller-turnaround Op Li ,mask= Ns Ar qualifier Count memory control turnaround events. This event may be further qualified using .Ar qualifier , which is a .Ql + separated set of the following keywords:
p
l -tag -width indent -compact XXX doc is unclear whether these are cycle counts or event counts
t Li dimm-turnaround Count DIMM turnarounds.
t Li read-to-write-turnaround Count read to write turnarounds.
t Li write-to-read-turnaround Count write to read turnarounds.
.El
p The default is to count all types of events. t Li k8-nb-ht-bus0-bandwidth Op Li ,mask= Ns Ar qualifier t Li k8-nb-ht-bus1-bandwidth Op Li ,mask= Ns Ar qualifier t Li k8-nb-ht-bus2-bandwidth Op Li ,mask= Ns Ar qualifier Count events on the HyperTransport(tm) buses. These events may be further qualified using .Ar qualifier , which is a .Ql + separated set of the following keywords:
p l -tag -width indent -compact t Li buffer-release Count buffer release messages sent. t Li command Count command messages sent. t Li data Count data messages sent. t Li nop Count nop messages sent. .El
p The default is to count all types of messages. .El .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. l -column "branch-mispredicts" "Description" t Em Alias Ta Em Event t Li branches Ta Li k8-fr-retired-taken-branches t Li branch-mispredicts Ta Li k8-fr-retired-taken-branches-mispredicted t Li dc-misses Ta Li k8-dc-miss t Li ic-misses Ta Li k8-ic-miss t Li instructions Ta Li k8-fr-retired-x86-instructions t Li interrupts Ta Li k8-fr-taken-hardware-interrupts t Li unhalted-cycles Ta Li k8-by-cpu-clk-unhalted .El .Sh SEE ALSO .Xr pmc 3 , .Xr pmc.k7 3 , .Xr pmc.p4 3 , .Xr pmc.p5 3 , .Xr pmc.p6 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 .