xref: /freebsd/lib/libpmc/pmc.atom.3 (revision dcc3a33188bceb5b6e819efdb9c5f72d059084b6)
1.\" Copyright (c) 2008 Joseph Koshy.  All rights reserved.
2.\"
3.\" Redistribution and use in source and binary forms, with or without
4.\" modification, are permitted provided that the following conditions
5.\" are met:
6.\" 1. Redistributions of source code must retain the above copyright
7.\"    notice, this list of conditions and the following disclaimer.
8.\" 2. Redistributions in binary form must reproduce the above copyright
9.\"    notice, this list of conditions and the following disclaimer in the
10.\"    documentation and/or other materials provided with the distribution.
11.\"
12.\" This software is provided by Joseph Koshy ``as is'' and
13.\" any express or implied warranties, including, but not limited to, the
14.\" implied warranties of merchantability and fitness for a particular purpose
15.\" are disclaimed.  in no event shall Joseph Koshy be liable
16.\" for any direct, indirect, incidental, special, exemplary, or consequential
17.\" damages (including, but not limited to, procurement of substitute goods
18.\" or services; loss of use, data, or profits; or business interruption)
19.\" however caused and on any theory of liability, whether in contract, strict
20.\" liability, or tort (including negligence or otherwise) arising in any way
21.\" out of the use of this software, even if advised of the possibility of
22.\" such damage.
23.\"
24.\" $FreeBSD$
25.\"
26.Dd November 12, 2008
27.Os
28.Dt PMC.ATOM 3
29.Sh NAME
30.Nm pmc.atom
31.Nd measurement events for
32.Tn Intel
33.Tn Atom
34family CPUs
35.Sh LIBRARY
36.Lb libpmc
37.Sh SYNOPSIS
38.In pmc.h
39.Sh DESCRIPTION
40.Tn Intel
41.Tn Atom
42CPUs contain PMCs conforming to version 3 of the
43.Tn Intel
44performance measurement architecture.
45These CPUs contains two classes of PMCs:
46.Bl -tag -width "Li PMC_CLASS_IAP"
47.It Li PMC_CLASS_IAF
48Fixed-function counters that count only one hardware event per counter.
49.It Li PMC_CLASS_IAP
50Programmable counters that may be configured to count one of a defined
51set of hardware events.
52.El
53.Pp
54The number of PMCs available in each class and their widths need to be
55determined at run time by calling
56.Xr pmc_cpuinfo 3 .
57.Pp
58Intel Atom PMCs are documented in
59.Rs
60.%B "IA-32 Intel(R) Architecture Software Developer's Manual"
61.%T "Volume 3: System Programming Guide"
62.%N "Order Number 253669-027US"
63.%D July 2008
64.%Q "Intel Corporation"
65.Re
66.Ss ATOM FIXED FUNCTION PMCS
67These PMCs and their supported events are documented in
68.Xr pmc.iaf 3 .
69.Ss ATOM PROGRAMMABLE PMCS
70The programmable PMCs support the following capabilities:
71.Bl -column "PMC_CAP_INTERRUPT" "Support"
72.It Em Capability Ta Em Support
73.It PMC_CAP_CASCADE Ta \&No
74.It PMC_CAP_EDGE Ta Yes
75.It PMC_CAP_INTERRUPT Ta Yes
76.It PMC_CAP_INVERT Ta Yes
77.It PMC_CAP_READ Ta Yes
78.It PMC_CAP_PRECISE Ta \&No
79.It PMC_CAP_SYSTEM Ta Yes
80.It PMC_CAP_TAGGING Ta \&No
81.It PMC_CAP_THRESHOLD Ta Yes
82.It PMC_CAP_USER Ta Yes
83.It PMC_CAP_WRITE Ta Yes
84.El
85.Ss Event Qualifiers
86Event specifiers for these PMCs support the following common
87qualifiers:
88.Bl -tag -width indent
89.It Li any
90Count matching events seen on any logical processor in a package.
91.It Li cmask= Ns Ar value
92Configure the PMC to increment only if the number of configured
93events measured in a cycle is greater than or equal to
94.Ar value .
95.It Li edge
96Configure the PMC to count the number of de-asserted to asserted
97transitions of the conditions expressed by the other qualifiers.
98If specified, the counter will increment only once whenever a
99condition becomes true, irrespective of the number of clocks during
100which the condition remains true.
101.It Li inv
102Invert the sense of comparison when the
103.Dq Li cmask
104qualifier is present, making the counter increment when the number of
105events per cycle is less than the value specified by the
106.Dq Li cmask
107qualifier.
108.It Li os
109Configure the PMC to count events happening at processor privilege
110level 0.
111.It Li usr
112Configure the PMC to count events occurring at privilege levels 1, 2
113or 3.
114.El
115.Pp
116If neither of the
117.Dq Li os
118or
119.Dq Li usr
120qualifiers are specified, the default is to enable both.
121.Pp
122Events that require core-specificity to be specified use a
123additional qualifier
124.Dq Li core= Ns Ar core ,
125where argument
126.Ar core
127is one of:
128.Bl -tag -width indent
129.It Li all
130Measure event conditions on all cores.
131.It Li this
132Measure event conditions on this core.
133.El
134.Pp
135The default is
136.Dq Li this .
137.Pp
138Events that require an agent qualifier to be specified use an
139additional qualifier
140.Dq Li agent= Ns agent ,
141where argument
142.Ar agent
143is one of:
144.Bl -tag -width indent
145.It Li this
146Measure events associated with this bus agent.
147.It Li any
148Measure events caused by any bus agent.
149.El
150.Pp
151The default is
152.Dq Li this .
153.Pp
154Events that require a hardware prefetch qualifier to be specified use an
155additional qualifier
156.Dq Li prefetch= Ns Ar prefetch ,
157where argument
158.Ar prefetch
159is one of:
160.Bl -tag -width "exclude"
161.It Li both
162Include all prefetches.
163.It Li only
164Only count hardware prefetches.
165.It Li exclude
166Exclude hardware prefetches.
167.El
168.Pp
169The default is
170.Dq Li both .
171.Pp
172Events that require a cache coherence qualifier to be specified use an
173additional qualifier
174.Dq Li cachestate= Ns Ar state ,
175where argument
176.Ar state
177contains one or more of the following letters:
178.Bl -tag -width indent
179.It Li e
180Count cache lines in the exclusive state.
181.It Li i
182Count cache lines in the invalid state.
183.It Li m
184Count cache lines in the modified state.
185.It Li s
186Count cache lines in the shared state.
187.El
188.Pp
189The default is
190.Dq Li eims .
191.Pp
192Events that require a snoop response qualifier to be specified use an
193additional qualifier
194.Dq Li snoopresponse= Ns Ar response ,
195where argument
196.Ar response
197comprises of the following keywords separated by
198.Dq +
199signs:
200.Bl -tag -width indent
201.It Li clean
202Measure CLEAN responses.
203.It Li hit
204Measure HIT responses.
205.It Li hitm
206Measure HITM responses.
207.El
208.Pp
209The default is to measure all the above responses.
210.Pp
211Events that require a snoop type qualifier use an additional qualifier
212.Dq Li snooptype= Ns Ar type ,
213where argument
214.Ar type
215comprises the one of the following keywords:
216.Bl -tag -width indent
217.It Li cmp2i
218Measure CMP2I snoops.
219.It Li cmp2s
220Measure CMP2S snoops.
221.El
222.Pp
223The default is to measure both snoops.
224.Ss Event Specifiers (Programmable PMCs)
225Atom programmable PMCs support the following events:
226.Bl -tag -width indent
227.It Li BACLEARS
228.Pq Event E6H , Umask 01H
229The number of times the front end is resteered.
230.It Li BOGUS_BR
231.Pq Event E4H , Umask 00H
232The number of byte sequences mistakenly detected as taken branch
233instructions.
234.It Li BR_BAC_MISSP_EXEC
235.Pq Event 8AH , Umask 00H
236The number of branch instructions that were mispredicted when
237decoded.
238.It Li BR_CALL_MISSP_EXEC
239.Pq Event 93H , Umask 00H
240The number of mispredicted
241.Li CALL
242instructions that were executed.
243.It Li BR_CALL_EXEC
244.Pq Event 92H , Umask 00H
245The number of
246.Li CALL
247instructions executed.
248.It Li BR_CND_EXEC
249.Pq Event 8BH , Umask 00H
250The number of conditional branches executed, but not necessarily retired.
251.It Li BR_CND_MISSP_EXEC
252.Pq Event 8CH , Umask 00H
253The number of mispredicted conditional branches executed.
254.It Li BR_IND_CALL_EXEC
255.Pq Event 94H , Umask 00H
256The number of indirect
257.Li CALL
258instructions executed.
259.It Li BR_IND_EXEC
260.Pq Event 8DH , Umask 00H
261The number of indirect branch instructions executed.
262.It Li BR_IND_MISSP_EXEC
263.Pq Event 8EH , Umask 00H
264The number of mispredicted indirect branch instructions executed.
265.It Li BR_INST_DECODED
266.Pq Event E0H , Umask 01H
267The number of branch instructions decoded.
268.It Li BR_INST_EXEC
269.Pq Event 88H , Umask 00H
270The number of branches executed, but not necessarily retired.
271.It Li BR_INST_RETIRED.ANY
272.Pq Event C4H , Umask 00H
273.Pq Alias Qq "Branch Instruction Retired"
274The number of branch instructions retired.
275This is an architectural performance event.
276.It Li BR_INST_RETIRED.ANY1
277.Pq Event C4H , Umask 0FH
278The number of branch instructions retired that were mispredicted.
279.It Li BR_INST_RETIRED.MISPRED
280.Pq Event C5H , Umask 00H
281.Pq Alias Qq "Branch Misses Retired"
282The number of mispredicted branch instructions retired.
283This is an architectural performance event.
284.It Li BR_INST_RETIRED.MISPRED_NOT_TAKEN
285.Pq Event C4H , Umask 02H
286The number of not taken branch instructions retired that were
287mispredicted.
288.It Li BR_INST_RETIRED.MISPRED_TAKEN
289.Pq Event C4H , Umask 08H
290The number taken branch instructions retired that were mispredicted.
291.It Li BR_INST_RETIRED.PRED_NOT_TAKEN
292.Pq Event C4H , Umask 01H
293The number of not taken branch instructions retired that were
294correctly predicted.
295.It Li BR_INST_RETIRED.PRED_TAKEN
296.Pq Event C4H , Umask 04H
297The number of taken branch instructions retired that were correctly
298predicted.
299.It Li BR_INST_RETIRED.TAKEN
300.Pq Event C4H , Umask 0CH
301The number of taken branch instructions retired.
302.It Li BR_MISSP_EXEC
303.Pq Event 89H , Umask 00H
304The number of mispredicted branch instructions that were executed.
305.It Li BR_RET_MISSP_EXEC
306.Pq Event 90H , Umask 00H
307The number of mispredicted
308.Li RET
309instructions executed.
310.It Li BR_RET_BAC_MISSP_EXEC
311.Pq Event 91H , Umask 00H
312The number of
313.Li RET
314instructions executed that were mispredicted at decode time.
315.It Li BR_RET_EXEC
316.Pq Event 8FH , Umask 00H
317The number of
318.Li RET
319instructions executed.
320.It Li BR_TKN_BUBBLE_1
321.Pq Event 97H , Umask 00H
322The number of branch predicted taken with bubble 1.
323.It Li BR_TKN_BUBBLE_2
324.Pq Event 98H , Umask 00H
325The number of branch predicted taken with bubble 2.
326.It Li BUSQ_EMPTY Op ,core= Ns Ar core
327.Pq Event 7DH
328The number of cycles during which the core did not have any pending
329transactions in the bus queue.
330.It Li BUS_BNR_DRV Op ,agent= Ns Ar agent
331.Pq Event 61H
332The number of Bus Not Ready signals asserted on the bus.
333This event is thread-independent.
334.It Li BUS_DATA_RCV Op ,core= Ns Ar core
335.Pq Event 64H
336The number of bus cycles during which the processor is receiving data.
337This event is thread-independent.
338.It Li BUS_DRDY_CLOCKS Op ,agent= Ns Ar agent
339.Pq Event 62H
340The number of bus cycles during which the Data Ready signal is asserted
341on the bus.
342This event is thread-independent.
343.It Li BUS_HIT_DRV Op ,agent= Ns Ar agent
344.Pq Event 7AH
345The number of bus cycles during which the processor drives the
346.Li HIT#
347pin.
348This event is thread-independent.
349.It Li BUS_HITM_DRV Op ,agent= Ns Ar agent
350.Pq Event 7BH
351The number of bus cycles during which the processor drives the
352.Li HITM#
353pin.
354This event is thread-independent.
355.It Li BUS_IO_WAIT Op ,core= Ns Ar core
356.Pq Event 7FH
357The number of core cycles during which I/O requests wait in the bus
358queue.
359.It Li BUS_LOCK_CLOCKS Xo
360.Op ,agent= Ns Ar agent
361.Op ,core= Ns Ar core
362.Xc
363.Pq Event 63H
364The number of bus cycles during which the
365.Li LOCK
366signal was asserted on the bus.
367This event is thread independent.
368.It Li BUS_REQUEST_OUTSTANDING Xo
369.Op ,agent= Ns Ar agent
370.Op ,core= Ns Ar core
371.Xc
372.Pq Event 60H
373The number of pending full cache line read transactions on the bus
374occurring in each cycle.
375This event is thread independent.
376.It Li BUS_TRANS_P Xo
377.Op ,agent= Ns Ar agent
378.Op ,core= Ns Ar core
379.Xc
380.Pq Event 6BH
381The number of partial bus transactions.
382.It Li BUS_TRANS_IFETCH Xo
383.Op ,agent= Ns Ar agent
384.Op ,core= Ns Ar core
385.Xc
386.Pq Event 68H
387The number of instruction fetch full cache line bus transactions.
388.It Li BUS_TRANS_INVAL Xo
389.Op ,agent= Ns Ar agent
390.Op ,core= Ns Ar core
391.Xc
392.Pq Event 69H
393The number of invalidate bus transactions.
394.It Li BUS_TRANS_PWR Xo
395.Op ,agent= Ns Ar agent
396.Op ,core= Ns Ar core
397.Xc
398.Pq Event 6AH
399The number of partial write bus transactions.
400.It Li BUS_TRANS_DEF Xo
401.Op ,agent= Ns Ar agent
402.Op ,core= Ns Ar core
403.Xc
404.Pq Event 6DH
405The number of deferred bus transactions.
406.It Li BUS_TRANS_BURST Xo
407.Op ,agent= Ns Ar agent
408.Op ,core= Ns Ar core
409.Xc
410.Pq Event 6EH
411The number of burst transactions.
412.It Li BUS_TRANS_MEM Xo
413.Op ,agent= Ns Ar agent
414.Op ,core= Ns Ar core
415.Xc
416.Pq Event 6FH
417The number of memory bus transactions.
418.It Li BUS_TRANS_ANY Xo
419.Op ,agent= Ns Ar agent
420.Op ,core= Ns Ar core
421.Xc
422.Pq Event 70H
423The number of bus transactions of any kind.
424.It Li BUS_TRANS_BRD Xo
425.Op ,agent= Ns Ar agent
426.Op ,core= Ns Ar core
427.Xc
428.Pq Event 65H
429The number of burst read transactions.
430.It Li BUS_TRANS_IO Xo
431.Op ,agent= Ns Ar agent
432.Op ,core= Ns Ar core
433.Xc
434.Pq Event 6CH
435The number of completed I/O bus transactions due to
436.Li IN
437and
438.Li OUT
439instructions.
440.It Li BUS_TRANS_RFO Xo
441.Op ,agent= Ns Ar agent
442.Op ,core= Ns Ar core
443.Xc
444.Pq Event 66H
445The number of Read For Ownership bus transactions.
446.It Li BUS_TRANS_WB Xo
447.Op ,agent= Ns Ar agent
448.Op ,core= Ns Ar core
449.Xc
450.Pq Event 67H
451The number explicit write-back bus transactions due to dirty line
452evictions.
453.It Li CMP_SNOOP Xo
454.Op ,core= Ns Ar core
455.Op ,snooptype= Ns Ar snoop
456.Xc
457.Pq Event 78H
458The number of times the L1 data cache is snooped by the other core in
459the same processor.
460.It Li CPU_CLK_UNHALTED.BUS
461.Pq Event 3CH , Umask 01H
462.Pq Alias Qq "Unhalted Reference Cycles"
463The number of bus cycles when the core is not in the halt state.
464This is an architectural performance event.
465.It Li CPU_CLK_UNHALTED.CORE_P
466.Pq Event 3CH , Umask 00H
467.Pq Alias Qq "Unhalted Core Cycles"
468The number of core cycles while the core is not in a halt state.
469This is an architectural performance event.
470.It Li CPU_CLK_UNHALTED.NO_OTHER
471.Pq Event 3CH , Umask 02H
472The number of bus cycles during which the core remains unhalted and
473the other core is halted.
474.It Li CYCLES_DIV_BUSY
475.Pq Event 14H , Umask 01H
476The number of cycles the divider is busy.
477.It Li CYCLES_INT_MASKED.CYCLES_INT_MASKED
478.Pq Event C6H , Umask 01H
479The number of cycles during which interrupts are disabled.
480.It Li CYCLES_INT_MASKED.CYCLES_INT_PENDING_AND_MASKED
481.Pq Event C6H , Umask 02H
482The number of cycles during which there were pending interrupts while
483interrupts were disabled.
484.It Li CYCLES_L1I_MEM_STALLED
485.Pq Event 86H , Umask 00H
486The number of cycles for which an instruction fetch stalls.
487.It Li DATA_TLB_MISSES.DTLB_MISS
488.Pq Event 08H , Umask 07H
489The number of memory access that missed the Data TLB
490.It Li DATA_TLB_MISSES.DTLB_MISS_LD
491.Pq Event 08H , Umask 05H
492The number of loads that missed the Data TLB.
493.It Li DATA_TLB_MISSES.DTLB_MISS_ST
494.Pq Event 08H , Umask 06H
495The number of stores that missed the Data TLB.
496.It Li DATA_TLB_MISSES.UTLB_MISS_LD
497.Pq Event 08H , Umask 09H
498The number of loads that missed the UTLB.
499.It Li DELAYED_BYPASS.FP
500.Pq Event 19H , Umask 00H
501The number of floating point operations that used data immediately
502after the data was generated by a non floating point execution unit.
503.It Li DELAYED_BYPASS.LOAD
504.Pq Event 19H , Umask 01H
505The number of delayed bypass penalty cycles that a load operation incurred.
506.It Li DELAYED_BYPASS.SIMD
507.Pq Event 19H , Umask 02H
508The number of times SIMD operations use data immediately after data,
509was generated by a non-SIMD execution unit.
510.It Li DIV
511.Pq Event 13H , Umask 00H
512The number of divide operations executed.
513This event is only available on PMC1.
514.It Li DIV.AR
515.Pq Event 13H , Umask 81H
516The number of divide operations retired.
517.It Li DIV.S
518.Pq Event 13H , Umask 01H
519The number of divide operations executed.
520.It Li DTLB_MISSES.ANY
521.Pq Event 08H , Umask 01H
522The number of Data TLB misses, including misses that result from
523speculative accesses.
524.It Li DTLB_MISSES.L0_MISS_LD
525.Pq Event 08H , Umask 04H
526The number of level 0 DTLB misses due to load operations.
527.It Li DTLB_MISSES.MISS_LD
528.Pq Event 08H , Umask 02H
529The number of Data TLB misses due to load operations.
530.It Li DTLB_MISSES.MISS_ST
531.Pq Event 08H , Umask 08H
532The number of Data TLB misses due to store operations.
533.It Li EIST_TRANS
534.Pq Event 3AH , Umask 00H
535The number of Enhanced Intel SpeedStep Technology transitions.
536.It Li ESP.ADDITIONS
537.Pq Event ABH , Umask 02H
538The number of automatic additions to the
539.Li %esp
540register.
541.It Li ESP.SYNCH
542.Pq Event ABH , Umask 01H
543The number of times the
544.Li %esp
545register was explicitly used in an address expression after
546it is implicitly used by a
547.Li PUSH
548or
549.Li POP
550instruction.
551.It Li EXT_SNOOP Xo
552.Op ,agent= Ns Ar agent
553.Op ,snoopresponse= Ns Ar response
554.Xc
555.Pq Event 77H
556The number of snoop responses to bus transactions.
557.It Li FP_ASSIST
558.Pq Event 11H , Umask 01H
559The number of floating point operations executed that needed
560a microcode assist, including speculatively executed instructions.
561.It Li FP_ASSIST.AR
562.Pq Event 11H , Umask 81H
563The number of floating point operations retired that needed
564a microcode assist.
565.It Li FP_COMP_OPS_EXE
566.Pq Event 10H , Umask 00H
567The number of floating point computational micro-ops executed.
568The event is available only on PMC0.
569.It Li FP_MMX_TRANS_TO_FP
570.Pq Event CCH , Umask 02H
571The number of transitions from MMX instructions to floating point
572instructions.
573.It Li FP_MMX_TRANS_TO_MMX
574.Pq Event CCH , Umask 01H
575The number of transitions from floating point instructions to MMX
576instructions.
577.It Li HW_INT_RCV
578.Pq Event C8H , Umask 00H
579The number of hardware interrupts received.
580.It Li ICACHE.ACCESSES
581.Pq Event 80H , Umask 03H
582The number of instruction fetches.
583.It Li ICACHE.MISSES
584.Pq Event 80H , Umask 02H
585The number of instruction fetches that miss the instruction cache.
586.It Li IDLE_DURING_DIV
587.Pq Event 18H , Umask 00H
588The number of cycles the divider is busy and no other execution unit
589or load operation was in progress.
590This event is available only on PMC0.
591.It Li ILD_STALL
592.Pq Event 87H , Umask 00H
593The number of cycles the instruction length decoder stalled due to a
594length changing prefix.
595.It Li INST_QUEUE.FULL
596.Pq Event 83H , Umask 02H
597The number of cycles during which the instruction queue is full.
598.It Li INST_RETIRED.ANY_P
599.Pq Event C0H , Umask 00H
600.Pq Alias Qq "Instruction Retired"
601The number of instructions retired.
602This is an architectural performance event.
603.It Li INST_RETIRED.LOADS
604.Pq Event C0H , Umask 01H
605The number of instructions retired that contained a load operation.
606.It Li INST_RETIRED.OTHER
607.Pq Event C0H , Umask 04H
608The number of instructions retired that did not contain a load or a
609store operation.
610.It Li INST_RETIRED.STORES
611.Pq Event C0H , Umask 02H
612The number of instructions retired that contained a store operation.
613.It Li ITLB.FLUSH
614.Pq Event 82H , Umask 04H
615The number of ITLB flushes.
616.It Li ITLB.LARGE_MISS
617.Pq Event 82H , Umask 10H
618The number of instruction fetches from large pages that miss the
619ITLB.
620.It Li ITLB.MISSES
621.Pq Event 82H , Umask 02H
622The number of instruction fetches from both large and small pages that
623miss the ITLB.
624.It Li ITLB.SMALL_MISS
625.Pq Event 82H , Umask 02H
626The number of instruction fetches from small pages that miss the ITLB.
627.It Li ITLB_MISS_RETIRED
628.Pq Event C9H , Umask 00H
629The number of retired instructions that missed the ITLB when they were
630fetched.
631.It Li L1D_ALL_REF
632.Pq Event 43H , Umask 01H
633The number of references to L1 data cache counting loads and stores of
634to all memory types.
635.It Li L1D_ALL_CACHE_REF
636.Pq Event 43H , Umask 02H
637The number of data reads and writes to cacheable memory.
638.It Li L1D_CACHE_LOCK Op ,cachestate= Ns Ar state
639.Pq Event 42H
640The number of locked reads from cacheable memory.
641.It Li L1D_CACHE_LOCK_DURATION
642.Pq Event 42H , Umask 10H
643The number of cycles during which any cache line is locked by any
644locking instruction.
645.It Li L1D_CACHE.LD
646.Pq Event 40H , Umask 21H
647The number of data reads from cacheable memory.
648.It Li L1D_CACHE.ST
649.Pq Event 41H , Umask 22H
650The number of data writes to cacheable memory.
651.It Li L1D_M_EVICT
652.Pq Event 47H , Umask 00H
653The number of modified cache lines evicted from L1 data cache.
654.It Li L1D_M_REPL
655.Pq Event 46H , Umask 00H
656The number of modified lines allocated in L1 data cache.
657.It Li L1D_PEND_MISS
658.Pq Event 48H , Umask 00H
659The total number of outstanding L1 data cache misses at any clock.
660.It Li L1D_PREFETCH.REQUESTS
661.Pq Event 4EH , Umask 10H
662The number of times L1 data cache requested to prefetch a data cache
663line.
664.It Li L1D_REPL
665.Pq Event 45H , Umask 0FH
666The number of lines brought into L1 data cache.
667.It Li L1D_SPLIT.LOADS
668.Pq Event 49H , Umask 01H
669The number of load operations that span two cache lines.
670.It Li L1D_SPLIT.STORES
671.Pq Event 49H , Umask 02H
672The number of store operations that span two cache lines.
673.It Li L1I_MISSES
674.Pq Event 81H , Umask 00H
675The number of instruction fetch unit misses.
676.It Li L1I_READS
677.Pq Event 80H , Umask 00H
678The number of instruction fetches.
679.It Li L2_ADS Op ,core= Ns core
680.Pq Event 21H
681The number of cycles that the L2 address bus is in use.
682.It Li L2_DBUS_BUSY_RD Op ,core= Ns core
683.Pq Event 23H
684The number of core cycles during which the L2 data bus is busy
685transferring data to the core.
686.It Li L2_IFETCH Xo
687.Op ,cachestate= Ns Ar state
688.Op ,core= Ns Ar core
689.Xc
690.Pq Event 28H
691The number of instruction cache line requests from the instruction
692fetch unit.
693.It Li L2_LD Xo
694.Op ,cachestate= Ns Ar state
695.Op ,core= Ns Ar core
696.Op ,prefetch= Ns Ar prefetch
697.Xc
698.Pq Event 29H
699The number of L2 cache read requests from L1 cache and L2
700prefetchers.
701.It Li L2_LINES_IN Xo
702.Op ,core= Ns Ar core
703.Op ,prefetch= Ns Ar prefetch
704.Xc
705.Pq Event 24H
706The number of cache lines allocated in L2 cache.
707.It Li L2_LINES_OUT Xo
708.Op ,core= Ns Ar core
709.Op ,prefetch= Ns Ar prefetch
710.Xc
711.Pq Event 26H
712The number of L2 cache lines evicted.
713.It Li L2_LOCK Xo
714.Op ,cachestate= Ns Ar state
715.Op ,core= Ns Ar core
716.Xc
717.Pq Event 2BH
718The number of locked accesses to cache lines that miss L1 data
719cache.
720.It Li L2_M_LINES_IN Op ,core= Ns Ar core
721.Pq Event 25H
722The number of L2 cache line modifications.
723.It Li L2_M_LINES_OUT Xo
724.Op ,core= Ns Ar core
725.Op ,prefetch= Ns Ar prefetch
726.Xc
727.Pq Event 27H
728The number of modified lines evicted from L2 cache.
729.It Li L2_NO_REQ Op ,core= Ns Ar core
730.Pq Event 32H
731The number of cycles during which no L2 cache requests were pending
732from a core.
733.It Li L2_REJECT_BUSQ Xo
734.Op ,cachestate= Ns Ar state
735.Op ,core= Ns Ar core
736.Op ,prefetch= Ns Ar prefetch
737.Xc
738.Pq Event 30H
739The number of L2 cache requests that were rejected.
740.It Li L2_RQSTS Xo
741.Op ,cachestate= Ns Ar state
742.Op ,core= Ns Ar core
743.Op ,prefetch= Ns Ar prefetch
744.Xc
745.Pq Event 2EH
746The number of completed L2 cache requests.
747.It Li L2_RQSTS.SELF.DEMAND.I_STATE
748.Pq Event 2EH , Umask 41H
749.Pq Alias Qq "LLC Misses"
750The number of completed L2 cache demand requests from this core that
751missed the L2 cache.
752This is an architectural performance event.
753.It Li L2_RQSTS.SELF.DEMAND.MESI
754.Pq Event 2EH , Umask 4FH
755.Pq Alias Qq "LLC References"
756The number of completed L2 cache demand requests from this core.
757.It Li L2_ST Xo
758.Op ,cachestate= Ns Ar state
759.Op ,core= Ns Ar core
760.Xc
761.Pq Event 2AH
762The number of store operations that miss the L1 cache and request data
763from the L2 cache.
764.It Li LOAD_BLOCK.L1D
765.Pq Event 03H , Umask 20H
766The number of loads blocked by the L1 data cache.
767.It Li LOAD_BLOCK.OVERLAP_STORE
768.Pq Event 03H , Umask 08H
769The number of loads that partially overlap an earlier store or are
770aliased with a previous store.
771.It Li LOAD_BLOCK.STA
772.Pq Event 03H , Umask 02H
773The number of loads blocked by preceding stores whose address is yet
774to be calculated.
775.It Li LOAD_BLOCK.STD
776.Pq Event 03H , Umask 04H
777The number of loads blocked by preceding stores to the same address
778whose data value is not known.
779.It Li LOAD_BLOCK.UNTIL_RETIRE
780.Pq Event 03H , Umask 10H
781The number of load operations that were blocked until retirement.
782.It Li LOAD_HIT_PRE
783.Pq Event 4CH , Umask 00H
784The number of load operations that conflicted with an prefetch to the
785same cache line.
786.It Li MACHINE_CLEARS.SMC
787.Pq Event C3H , Umask 01H
788The number of times a program writes to a code section.
789.It Li MACHINE_NUKES.MEM_ORDER
790.Pq Event C3H , Umask 04H
791The number of times the execution pipeline was restarted due to a
792memory ordering conflict or memory disambiguation misprediction.
793.It Li MACRO_INSTS.ALL_DECODED
794.Pq Event AAH , Umask 03H
795The number of instructions decoded.
796.It Li MACRO_INSTS.CISC_DECODED
797.Pq Event AAH , Umask 02H
798The number of complex instructions decoded.
799.It Li MEMORY_DISAMBIGUATION.RESET
800.Pq Event 09H , Umask 01H
801The number of cycles during which memory disambiguation misprediction
802occurs.
803.It Li MEMORY_DISAMBIGUATION.SUCCESS
804.Pq Event 09H , Umask 02H
805The number of load operations that were successfully disambiguated.
806.It Li MEM_LOAD_RETIRED.DTLB_MISS
807.Pq Event CBH , Umask 04H
808The number of retired load operations that missed the DTLB.
809.It Li MEM_LOAD_RETIRED.L2_MISS
810.Pq Event CBH , Umask 02H
811The number of retired load operations that miss L2 cache.
812.It Li MEM_LOAD_RETIRED.L2_HIT
813.Pq Event CBH , Umask 01H
814The number of retired load operations that hit L2 cache.
815.It Li MEM_LOAD_RETIRED.L2_LINE_MISS
816.Pq Event CBH , Umask 08H
817The number of load operations that missed L2 cache and that caused a
818bus request.
819.It Li MUL
820.Pq Event 12H , Umask 00H
821The number of multiply operations executed.
822This event is only available on PMC1.
823.It Li MUL.AR
824.Pq Event 12H , Umask 81H
825The number of multiply operations retired.
826.It Li MUL.S
827.Pq Event 12H , Umask 01H
828The number of multiply operations executed.
829.It Li PAGE_WALKS.WALKS
830.Pq Event 0CH , Umask 03H
831The number of page walks executed due to an ITLB or DTLB miss.
832.It Li PAGE_WALKS.CYCLES
833.Pq Event 0CH , Umask 03H
834.\" XXX Clarify.  Identical event umask/event numbers.
835The number of cycles spent in a page walk caused by an ITLB or DTLB
836miss.
837.It Li PREF_RQSTS_DN
838.Pq Event F8H , Umask 00H
839The number of downward prefetches issued from the Data Prefetch Logic
840unit to L2 cache.
841.It Li PREF_RQSTS_UP
842.Pq Event F0H , Umask 00H
843The number of upward prefetches issued from the Data Prefetch Logic
844unit to L2 cache.
845.It Li PREFETCH.PREFETCHNTA
846.Pq Event 07H , Umask 08H
847The number of
848.Li PREFETCHNTA
849instructions executed.
850.It Li PREFETCH.PREFETCHT0
851.Pq Event 07H , Umask 01H
852The number of
853.Li PREFETCHT0
854instructions executed.
855.It Li PREFETCH.SW_L2
856.Pq Event 07H , Umask 06H
857The number of
858.Li PREFETCHT1
859and
860.Li PREFETCHT2
861instructions executed.
862.It Li RAT_STALLS.ANY
863.Pq Event D2H , Umask 0FH
864The number of stall cycles due to any of
865.Li RAT_STALLS.FLAGS
866.Li RAT_STALLS.FPSW ,
867.Li RAT_STALLS.PARTIAL
868and
869.Li RAT_STALLS.ROB_READ_PORT .
870.It Li RAT_STALLS.FLAGS
871.Pq Event D2H , Umask 04H
872The number of cycles execution stalled due to a flag register induced
873stall.
874.It Li RAT_STALLS.FPSW
875.Pq Event D2H , Umask 08H
876The number of times the floating point status word was written.
877.It Li RAT_STALLS.PARTIAL_CYCLES
878.Pq Event D2H , Umask 02H
879The number of cycles of added instruction execution latency due to the
880use of a register that was partially written by previous instructions.
881.It Li RAT_STALLS.ROB_READ_PORT
882.Pq Event D2H , Umask 01H
883The number of cycles when ROB read port stalls occurred.
884.It Li RESOURCE_STALLS.ANY
885.Pq Event DCH , Umask 1FH
886The number of cycles during which any resource related stall
887occurred.
888.It Li RESOURCE_STALLS.BR_MISS_CLEAR
889.Pq Event DCH , Umask 10H
890The number of cycles stalled due to branch misprediction.
891.It Li RESOURCE_STALLS.FPCW
892.Pq Event DCH , Umask 08H
893The number of cycles stalled due to writing the floating point control
894word.
895.It Li RESOURCE_STALLS.LD_ST
896.Pq Event DCH , Umask 04H
897The number of cycles during which the number of loads and stores in
898the pipeline exceeded their limits.
899.It Li RESOURCE_STALLS.ROB_FULL
900.Pq Event DCH , Umask 01H
901The number of cycles when the reorder buffer was full.
902.It Li RESOURCE_STALLS.RS_FULL
903.Pq Event DCH , Umask 02H
904The number of cycles during which the RS was full.
905.It Li RS_UOPS_DISPATCHED
906.Pq Event A0H , Umask 00H
907The number of micro-ops dispatched for execution.
908.It Li RS_UOPS_DISPATCHED.PORT0
909.Pq Event A1H , Umask 01H
910The number of cycles micro-ops were dispatched for execution on port
9110.
912.It Li RS_UOPS_DISPATCHED.PORT1
913.Pq Event A1H , Umask 02H
914The number of cycles micro-ops were dispatched for execution on port
9151.
916.It Li RS_UOPS_DISPATCHED.PORT2
917.Pq Event A1H , Umask 04H
918The number of cycles micro-ops were dispatched for execution on port
9192.
920.It Li RS_UOPS_DISPATCHED.PORT3
921.Pq Event A1H , Umask 08H
922The number of cycles micro-ops were dispatched for execution on port
9233.
924.It Li RS_UOPS_DISPATCHED.PORT4
925.Pq Event A1H , Umask 10H
926The number of cycles micro-ops were dispatched for execution on port
9274.
928.It Li RS_UOPS_DISPATCHED.PORT5
929.Pq Event A1H , Umask 20H
930The number of cycles micro-ops were dispatched for execution on port
9315.
932.It Li SB_DRAIN_CYCLES
933.Pq Event 04H , Umask 01H
934The number of cycles while the store buffer is draining.
935.It Li SEGMENT_REG_LOADS.ANY
936.Pq Event 06H , Umask 00H
937The number of segment register loads.
938.It Li SEG_REG_RENAMES.ANY
939.Pq Event D5H , Umask 0FH
940The number of times the any segment register was renamed.
941.It Li SEG_REG_RENAMES.DS
942.Pq Event D5H , Umask 02H
943The number of times the
944.Li %ds
945register is renamed.
946.It Li SEG_REG_RENAMES.ES
947.Pq Event D5H , Umask 01H
948The number of times the
949.Li %es
950register is renamed.
951.It Li SEG_REG_RENAMES.FS
952.Pq Event D5H , Umask 04H
953The number of times the
954.Li %fs
955register is renamed.
956.It Li SEG_REG_RENAMES.GS
957.Pq Event D5H , Umask 08H
958The number of times the
959.Li %gs
960register is renamed.
961.It Li SEG_RENAME_STALLS.ANY
962.Pq Event D4H , Umask 0FH
963The number of stalls due to lack of resource to rename any segment
964register.
965.It Li SEG_RENAME_STALLS.DS
966.Pq Event D4H , Umask 02H
967The number of stalls due to lack of renaming resources for the
968.Li %ds
969register.
970.It Li SEG_RENAME_STALLS.ES
971.Pq Event D4H , Umask 01H
972The number of stalls due to lack of renaming resources for the
973.Li %es
974register.
975.It Li SEG_RENAME_STALLS.FS
976.Pq Event D4H , Umask 04H
977The number of stalls due to lack of renaming resources for the
978.Li %fs
979register.
980.It Li SEG_RENAME_STALLS.GS
981.Pq Event D4H , Umask 08H
982The number of stalls due to lack of renaming resources for the
983.Li %gs
984register.
985.It Li SIMD_ASSIST
986.Pq Event CDH , Umask 00H
987The number SIMD assists invoked.
988.It Li SIMD_COMP_INST_RETIRED.PACKED_DOUBLE
989.Pq Event CAH , Umask 04H
990Then number of computational SSE2 packed double precision instructions
991retired.
992.It Li SIMD_COMP_INST_RETIRED.PACKED_SINGLE
993.Pq Event CAH , Umask 01H
994Then number of computational SSE2 packed single precision instructions
995retired.
996.It Li SIMD_COMP_INST_RETIRED.SCALAR_DOUBLE
997.Pq Event CAH , Umask 08H
998Then number of computational SSE2 scalar double precision instructions
999retired.
1000.It Li SIMD_COMP_INST_RETIRED.SCALAR_SINGLE
1001.Pq Event CAH , Umask 02H
1002Then number of computational SSE2 scalar single precision instructions
1003retired.
1004.It Li SIMD_INSTR_RETIRED
1005.Pq Event CEH , Umask 00H
1006The number of retired SIMD instructions that use MMX registers.
1007.It Li SIMD_INST_RETIRED.ANY
1008.Pq Event C7H , Umask 1FH
1009The number of streaming SIMD instructions retired.
1010.It Li SIMD_INST_RETIRED.PACKED_DOUBLE
1011.Pq Event C7H , Umask 04H
1012The number of SSE2 packed double precision instructions retired.
1013.It Li SIMD_INST_RETIRED.PACKED_SINGLE
1014.Pq Event C7H , Umask 01H
1015The number of SSE packed single precision instructions retired.
1016.It Li SIMD_INST_RETIRED.SCALAR_DOUBLE
1017.Pq Event C7H , Umask 08H
1018The number of SSE2 scalar double precision instructions retired.
1019.It Li SIMD_INST_RETIRED.SCALAR_SINGLE
1020.Pq Event C7H , Umask 02H
1021The number of SSE scalar single precision instructions retired.
1022.It Li SIMD_INST_RETIRED.VECTOR
1023.Pq Event C7H , Umask 10H
1024The number of SSE2 vector instructions retired.
1025.It Li SIMD_SAT_INSTR_RETIRED
1026.Pq Event CFH , Umask 00H
1027The number of saturated arithmetic SIMD instructions retired.
1028.It Li SIMD_SAT_UOP_EXEC.AR
1029.Pq Event B1H , Umask 80H
1030The number of SIMD saturated arithmetic micro-ops retired.
1031.It Li SIMD_SAT_UOP_EXEC.S
1032.Pq Event B1H , Umask 00H
1033The number of SIMD saturated arithmetic micro-ops executed.
1034.It Li SIMD_UOPS_EXEC.AR
1035.Pq Event B0H , Umask 80H
1036The number of SIMD micro-ops retired.
1037.It Li SIMD_UOPS_EXEC.S
1038.Pq Event B0H , Umask 00H
1039The number of SIMD micro-ops executed.
1040.It Li SIMD_UOP_TYPE_EXEC.ARITHMETIC.AR
1041.Pq Event B3H , Umask A0H
1042The number of SIMD packed arithmetic micro-ops executed.
1043.It Li SIMD_UOP_TYPE_EXEC.ARITHMETIC.S
1044.Pq Event B3H , Umask 20H
1045The number of SIMD packed arithmetic micro-ops executed.
1046.It Li SIMD_UOP_TYPE_EXEC.LOGICAL.AR
1047.Pq Event B3H , Umask 90H
1048The number of SIMD packed logical micro-ops executed.
1049.It Li SIMD_UOP_TYPE_EXEC.LOGICAL.S
1050.Pq Event B3H , Umask 10H
1051The number of SIMD packed logical micro-ops executed.
1052.It Li SIMD_UOP_TYPE_EXEC.MUL.AR
1053.Pq Event B3H , Umask 81H
1054The number of SIMD packed multiply micro-ops retired.
1055.It Li SIMD_UOP_TYPE_EXEC.MUL.S
1056.Pq Event B3H , Umask 01H
1057The number of SIMD packed multiply micro-ops executed.
1058.It Li SIMD_UOP_TYPE_EXEC.PACK.AR
1059.Pq Event B3H , Umask 84H
1060The number of SIMD pack micro-ops retired.
1061.It Li SIMD_UOP_TYPE_EXEC.PACK.S
1062.Pq Event B3H , Umask 04H
1063The number of SIMD pack micro-ops executed.
1064.It Li SIMD_UOP_TYPE_EXEC.SHIFT.AR
1065.Pq Event B3H , Umask 82H
1066The number of SIMD packed shift micro-ops retired.
1067.It Li SIMD_UOP_TYPE_EXEC.SHIFT.S
1068.Pq Event B3H , Umask 02H
1069The number of SIMD packed shift micro-ops executed.
1070.It Li SIMD_UOP_TYPE_EXEC.UNPACK.AR
1071.Pq Event B3H , Umask 88H
1072The number of SIMD unpack micro-ops executed.
1073.It Li SIMD_UOP_TYPE_EXEC.UNPACK.S
1074.Pq Event B3H , Umask 08H
1075The number of SIMD unpack micro-ops executed.
1076.It Li SNOOP_STALL_DRV Xo
1077.Op ,agent= Ns Ar agent
1078.Op ,core= Ns Ar core
1079.Xc
1080.Pq Event 7EH
1081The number of times the bus stalled for snoops.
1082This event is thread-independent.
1083.It Li SSE_PRE_EXEC.L2
1084.Pq Event 07H , Umask 02H
1085The number of
1086.Li PREFETCHT1
1087instructions executed.
1088.It Li SSE_PRE_EXEC.STORES
1089.Pq Event 07H , Umask 03H
1090The number of times SSE non-temporal store instructions were executed.
1091.It Li SSE_PRE_MISS.L1
1092.Pq Event 4BH , Umask 01H
1093The number of times the
1094.Li PREFETCHT0
1095instruction executed and missed all cache levels.
1096.It Li SSE_PRE_MISS.L2
1097.Pq Event 4BH , Umask 02H
1098The number of times the
1099.Li PREFETCHT1
1100instruction executed and missed all cache levels.
1101.It Li SSE_PRE_MISS.NTA
1102.Pq Event 4BH , Umask 00H
1103The number of times the
1104.Li PREFETCHNTA
1105instruction executed and missed all cache levels.
1106.It Li STORE_BLOCK.ORDER
1107.Pq Event 04H , Umask 02H
1108The number of cycles while a store was waiting for another store to be
1109globally observed.
1110.It Li STORE_BLOCK.SNOOP
1111.Pq Event 04H , Umask 08H
1112The number of cycles while a store was blocked due to a conflict with
1113an internal or external snoop.
1114.It Li STORE_FORWARDS.GOOD
1115.Pq Event 02H , Umask 81H
1116The number of times stored data was forwarded directly to a load.
1117.It Li THERMAL_TRIP
1118.Pq Event 3BH , Umask C0H
1119The number of thermal trips.
1120.It Li UOPS_RETIRED.LD_IND_BR
1121.Pq Event C2H , Umask 01H
1122The number of micro-ops retired that fused a load with another
1123operation.
1124.It Li UOPS_RETIRED.STD_STA
1125.Pq Event C2H , Umask 02H
1126The number of store address calculations that fused into one micro-op.
1127.It Li UOPS_RETIRED.MACRO_FUSION
1128.Pq Event C2H , Umask 04H
1129The number of times retired instruction pairs were fused into one
1130micro-op.
1131.It Li UOPS_RETIRED.FUSED
1132.Pq Event C2H , Umask 07H
1133The number of fused micro-ops retired.
1134.It Li UOPS_RETIRED.NON_FUSED
1135.Pq Event C2H , Umask 8H
1136The number of non-fused micro-ops retired.
1137.It Li UOPS_RETIRED.ANY
1138.Pq Event C2H , Umask 10H
1139The number of micro-ops retired.
1140.It Li X87_COMP_OPS_EXE.ANY.AR
1141.Pq Event 10H , Umask 81H
1142The number of x87 floating-point computational micro-ops retired.
1143.It Li X87_COMP_OPS_EXE.ANY.S
1144.Pq Event 10H , Umask 01H
1145The number of x87 floating-point computational micro-ops executed.
1146.It Li X87_OPS_RETIRED.ANY
1147.Pq Event C1H , Umask FEH
1148The number of floating point computational instructions retired.
1149.It Li X87_OPS_RETIRED.FXCH
1150.Pq Event C1H , Umask 01H
1151The number of
1152.Li FXCH
1153instructions retired.
1154.El
1155.Ss Event Name Aliases
1156The following table shows the mapping between the PMC-independent
1157aliases supported by
1158.Lb libpmc
1159and the underlying hardware events used on these CPUs.
1160.Bl -column "branch-mispredicts" "cpu_clk_unhalted.core_p" "PMC Class"
1161.It Em Alias Ta Em Event Ta Em PMC Class
1162.It Li branches Ta Li BR_INST_RETIRED.ANY Ta Li PMC_CLASS_IAP
1163.It Li branch-mispredicts Ta Li BR_INST_RETIRED.MISPRED Ta Li PMC_CLASS_IAP
1164.It Li ic-misses Ta Li ICACHE.MISSES Ta Li PMC_CLASS_IAP
1165.It Li instructions Ta Li INST_RETIRED.ANY_P Ta Li PMC_CLASS_IAF
1166.It Li interrupts Ta Li HW_INT_RCV Ta Li PMC_CLASS_IAP
1167.It Li unhalted-cycles Ta Li CPU_CLK_UNHALTED.CORE_P Ta Li PMC_CLASS_IAF
1168.El
1169.Sh SEE ALSO
1170.Xr pmc 3 ,
1171.Xr pmc.core 3 ,
1172.Xr pmc.core2 3 ,
1173.Xr pmc.iaf 3 ,
1174.Xr pmc.k7 3 ,
1175.Xr pmc.k8 3 ,
1176.Xr pmc.p4 3 ,
1177.Xr pmc.p5 3 ,
1178.Xr pmc.p6 3 ,
1179.Xr pmc.tsc 3 ,
1180.Xr pmc_cpuinfo 3 ,
1181.Xr pmclog 3 ,
1182.Xr hwpmc 4
1183.Sh HISTORY
1184The
1185.Nm pmc
1186library first appeared in
1187.Fx 6.0 .
1188.Sh AUTHORS
1189The
1190.Lb libpmc
1191library was written by
1192.An "Joseph Koshy"
1193.Aq jkoshy@FreeBSD.org .
1194