xref: /freebsd/lib/libpmc/pmc.westmere.3 (revision 4f1f4356f3012928b463f9ef1710fb908e48b1e2)
1.\" Copyright (c) 2010 Fabien Thomas.  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 March 24, 2010
27.Dt PMC.WESTMERE 3
28.Os
29.Sh NAME
30.Nm pmc.westmere
31.Nd measurement events for
32.Tn Intel
33.Tn Westmere
34family CPUs
35.Sh LIBRARY
36.Lb libpmc
37.Sh SYNOPSIS
38.In pmc.h
39.Sh DESCRIPTION
40.Tn Intel
41.Tn "Westmere"
42CPUs contain PMCs conforming to version 2 of the
43.Tn Intel
44performance measurement architecture.
45These CPUs may contain up to three 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 Westmere PMCs are documented in
59.Rs
60.%B "Intel(R) 64 and IA-32 Architectures Software Developes Manual"
61.%T "Volume 3B: System Programming Guide, Part 2"
62.%N "Order Number: 253669-033US"
63.%D December 2009
64.%Q "Intel Corporation"
65.Re
66.Ss WESTMERE FIXED FUNCTION PMCS
67These PMCs and their supported events are documented in
68.Xr pmc.iaf 3 .
69.Ss WESTMERE 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 rsp= Ns Ar value
90Configure the Off-core Response bits.
91.Bl -tag -width indent
92.It Li DMND_DATA_RD
93Counts the number of demand and DCU prefetch data reads of full
94and partial cachelines as well as demand data page table entry
95cacheline reads. Does not count L2 data read prefetches or
96instruction fetches.
97.It Li DMND_RFO
98Counts the number of demand and DCU prefetch reads for ownership
99(RFO) requests generated by a write to data cacheline. Does not
100count L2 RFO.
101.It Li DMND_IFETCH
102Counts the number of demand and DCU prefetch instruction cacheline
103reads. Does not count L2 code read prefetches.
104WB
105Counts the number of writeback (modified to exclusive) transactions.
106.It Li PF_DATA_RD
107Counts the number of data cacheline reads generated by L2 prefetchers.
108.It Li PF_RFO
109Counts the number of RFO requests generated by L2 prefetchers.
110.It Li PF_IFETCH
111Counts the number of code reads generated by L2 prefetchers.
112.It Li OTHER
113Counts one of the following transaction types, including L3 invalidate,
114I/O, full or partial writes, WC or non-temporal stores, CLFLUSH, Fences,
115lock, unlock, split lock.
116.It Li UNCORE_HIT
117L3 Hit: local or remote home requests that hit L3 cache in the uncore
118with no coherency actions required (snooping).
119.It Li OTHER_CORE_HIT_SNP
120L3 Hit: local or remote home requests that hit L3 cache in the uncore
121and was serviced by another core with a cross core snoop where no modified
122copies were found (clean).
123.It Li OTHER_CORE_HITM
124L3 Hit: local or remote home requests that hit L3 cache in the uncore
125and was serviced by another core with a cross core snoop where modified
126copies were found (HITM).
127.It Li REMOTE_CACHE_FWD
128L3 Miss: local homed requests that missed the L3 cache and was serviced
129by forwarded data following a cross package snoop where no modified
130copies found. (Remote home requests are not counted)
131.It Li REMOTE_DRAM
132L3 Miss: remote home requests that missed the L3 cache and were serviced
133by remote DRAM.
134.It Li LOCAL_DRAM
135L3 Miss: local home requests that missed the L3 cache and were serviced
136by local DRAM.
137.It Li NON_DRAM
138Non-DRAM requests that were serviced by IOH.
139.El
140.It Li cmask= Ns Ar value
141Configure the PMC to increment only if the number of configured
142events measured in a cycle is greater than or equal to
143.Ar value .
144.It Li edge
145Configure the PMC to count the number of de-asserted to asserted
146transitions of the conditions expressed by the other qualifiers.
147If specified, the counter will increment only once whenever a
148condition becomes true, irrespective of the number of clocks during
149which the condition remains true.
150.It Li inv
151Invert the sense of comparison when the
152.Dq Li cmask
153qualifier is present, making the counter increment when the number of
154events per cycle is less than the value specified by the
155.Dq Li cmask
156qualifier.
157.It Li os
158Configure the PMC to count events happening at processor privilege
159level 0.
160.It Li usr
161Configure the PMC to count events occurring at privilege levels 1, 2
162or 3.
163.El
164.Pp
165If neither of the
166.Dq Li os
167or
168.Dq Li usr
169qualifiers are specified, the default is to enable both.
170.Ss Event Specifiers (Programmable PMCs)
171Westmere programmable PMCs support the following events:
172.Bl -tag -width indent
173.It Li LOAD_BLOCK.OVERLAP_STORE
174.Pq Event 03H , Umask 02H
175Loads that partially overlap an earlier store
176.It Li SB_DRAIN.ANY
177.Pq Event 04H , Umask 07H
178All Store buffer stall cycles
179.It Li MISALIGN_MEMORY.STORE
180.Pq Event 05H , Umask 02H
181All store referenced with misaligned address
182.It Li STORE_BLOCKS.AT_RET
183.Pq Event 06H , Umask 04H
184Counts number of loads delayed with at-Retirement block code. The following
185loads need to be executed at retirement and wait for all senior stores on
186the same thread to be drained: load splitting across 4K boundary (page
187split), load accessing uncacheable (UC or USWC) memory, load lock, and load
188with page table in UC or USWC memory region.
189.It Li STORE_BLOCKS.L1D_BLOCK
190.Pq Event 06H , Umask 08H
191Cacheable loads delayed with L1D block code
192.It Li PARTIAL_ADDRESS_ALIAS
193.Pq Event 07H , Umask 01H
194Counts false dependency due to partial address aliasing
195.It Li DTLB_LOAD_MISSES.ANY
196.Pq Event 08H , Umask 01H
197Counts all load misses that cause a page walk
198.It Li DTLB_LOAD_MISSES.WALK_COMPLETED
199.Pq Event 08H , Umask 02H
200Counts number of completed page walks due to load miss in the STLB.
201.It Li DTLB_LOAD_MISSES.WALK_CYCLES
202.Pq Event 08H , Umask 04H
203Cycles PMH is busy with a page walk due to a load miss in the STLB.
204.It Li DTLB_LOAD_MISSES.STLB_HIT
205.Pq Event 08H , Umask 10H
206Number of cache load STLB hits
207.It Li DTLB_LOAD_MISSES.PDE_MISS
208.Pq Event 08H , Umask 20H
209Number of DTLB cache load misses where the low part of the linear to
210physical address translation was missed.
211.It Li MEM_INST_RETIRED.LOADS
212.Pq Event 0BH , Umask 01H
213Counts the number of instructions with an architecturally-visible store
214retired on the architected path.
215In conjunction with ld_lat facility
216.It Li MEM_INST_RETIRED.STORES
217.Pq Event 0BH , Umask 02H
218Counts the number of instructions with an architecturally-visible store
219retired on the architected path.
220In conjunction with ld_lat facility
221.It Li MEM_INST_RETIRED.LATENCY_ABOVE_THRESHOLD
222.Pq Event 0BH , Umask 10H
223Counts the number of instructions exceeding the latency specified with
224ld_lat facility.
225In conjunction with ld_lat facility
226.It Li MEM_STORE_RETIRED.DTLB_MISS
227.Pq Event 0CH , Umask 01H
228The event counts the number of retired stores that missed the DTLB. The DTLB
229miss is not counted if the store operation causes a fault. Does not counter
230prefetches. Counts both primary and secondary misses to the TLB
231.It Li UOPS_ISSUED.ANY
232.Pq Event 0EH , Umask 01H
233Counts the number of Uops issued by the Register Allocation Table to the
234Reservation Station, i.e. the UOPs issued from the front end to the back
235end.
236.It Li UOPS_ISSUED.STALLED_CYCLES
237.Pq Event 0EH , Umask 01H
238Counts the number of cycles no Uops issued by the Register Allocation Table
239to the Reservation Station, i.e. the UOPs issued from the front end to the
240back end.
241set invert=1, cmask = 1
242.It Li UOPS_ISSUED.FUSED
243.Pq Event 0EH , Umask 02H
244Counts the number of fused Uops that were issued from the Register
245Allocation Table to the Reservation Station.
246.It Li MEM_UNCORE_RETIRED.LOCAL_HITM
247.Pq Event 0FH , Umask 02H
248Load instructions retired that HIT modified data in sibling core (Precise
249Event)
250.It Li MEM_UNCORE_RETIRED.LOCAL_DRAM_AND_REMOTE_CACHE_HIT
251.Pq Event 0FH , Umask 08H
252Load instructions retired local dram and remote cache HIT data sources
253(Precise Event)
254.It Li MEM_UNCORE_RETIRED.LOCAL_DRAM
255.Pq Event 0FH , Umask 10H
256Load instructions retired with a data source of local DRAM or locally homed
257remote cache HITM (Precise Event)
258.It Li MEM_UNCORE_RETIRED.REMOTE_DRAM
259.Pq Event 0FH , Umask 20H
260Load instructions retired remote DRAM and remote home-remote cache HITM
261(Precise Event)
262.It Li MEM_UNCORE_RETIRED.UNCACHEABLE
263.Pq Event 0FH , Umask 80H
264Load instructions retired I/O (Precise Event)
265.It Li FP_COMP_OPS_EXE.X87
266.Pq Event 10H , Umask 01H
267Counts the number of FP Computational Uops Executed. The number of FADD,
268FSUB, FCOM, FMULs, integer MULsand IMULs, FDIVs, FPREMs, FSQRTS, integer
269DIVs, and IDIVs. This event does not distinguish an FADD used in the middle
270of a transcendental flow from a separate FADD instruction.
271.It Li FP_COMP_OPS_EXE.MMX
272.Pq Event 10H , Umask 02H
273Counts number of MMX Uops executed.
274.It Li FP_COMP_OPS_EXE.SSE_FP
275.Pq Event 10H , Umask 04H
276Counts number of SSE and SSE2 FP uops executed.
277.It Li FP_COMP_OPS_EXE.SSE2_INTEGER
278.Pq Event 10H , Umask 08H
279Counts number of SSE2 integer uops executed.
280.It Li FP_COMP_OPS_EXE.SSE_FP_PACKED
281.Pq Event 10H , Umask 10H
282Counts number of SSE FP packed uops executed.
283.It Li FP_COMP_OPS_EXE.SSE_FP_SCALAR
284.Pq Event 10H , Umask 20H
285Counts number of SSE FP scalar uops executed.
286.It Li FP_COMP_OPS_EXE.SSE_SINGLE_PRECISION
287.Pq Event 10H , Umask 40H
288Counts number of SSE* FP single precision uops executed.
289.It Li FP_COMP_OPS_EXE.SSE_DOUBLE_PRECISION
290.Pq Event 10H , Umask 80H
291Counts number of SSE* FP double precision uops executed.
292.It Li SIMD_INT_128.PACKED_MPY
293.Pq Event 12H , Umask 01H
294Counts number of 128 bit SIMD integer multiply operations.
295.It Li SIMD_INT_128.PACKED_SHIFT
296.Pq Event 12H , Umask 02H
297Counts number of 128 bit SIMD integer shift operations.
298.It Li SIMD_INT_128.PACK
299.Pq Event 12H , Umask 04H
300Counts number of 128 bit SIMD integer pack operations.
301.It Li SIMD_INT_128.UNPACK
302.Pq Event 12H , Umask 08H
303Counts number of 128 bit SIMD integer unpack operations.
304.It Li SIMD_INT_128.PACKED_LOGICAL
305.Pq Event 12H , Umask 10H
306Counts number of 128 bit SIMD integer logical operations.
307.It Li SIMD_INT_128.PACKED_ARITH
308.Pq Event 12H , Umask 20H
309Counts number of 128 bit SIMD integer arithmetic operations.
310.It Li SIMD_INT_128.SHUFFLE_MOVE
311.Pq Event 12H , Umask 40H
312Counts number of 128 bit SIMD integer shuffle and move operations.
313.It Li LOAD_DISPATCH.RS
314.Pq Event 13H , Umask 01H
315Counts number of loads dispatched from the Reservation Station that bypass
316the Memory Order Buffer.
317.It Li LOAD_DISPATCH.RS_DELAYED
318.Pq Event 13H , Umask 02H
319Counts the number of delayed RS dispatches at the stage latch. If an RS
320dispatch can not bypass to LB, it has another chance to dispatch from the
321one-cycle delayed staging latch before it is written into the LB.
322.It Li LOAD_DISPATCH.MOB
323.Pq Event 13H , Umask 04H
324Counts the number of loads dispatched from the Reservation Station to the
325Memory Order Buffer.
326.It Li LOAD_DISPATCH.ANY
327.Pq Event 13H , Umask 07H
328Counts all loads dispatched from the Reservation Station.
329.It Li ARITH.CYCLES_DIV_BUSY
330.Pq Event 14H , Umask 01H
331Counts the number of cycles the divider is busy executing divide or square
332root operations. The divide can be integer, X87 or Streaming SIMD Extensions
333(SSE). The square root operation can be either X87 or SSE.
334Set 'edge =1, invert=1, cmask=1' to count the number of divides.
335Count may be incorrect When SMT is on
336.It Li ARITH.MUL
337.Pq Event 14H , Umask 02H
338Counts the number of multiply operations executed. This includes integer as
339well as floating point multiply operations but excludes DPPS mul and MPSAD.
340Count may be incorrect When SMT is on
341.It Li INST_QUEUE_WRITES
342.Pq Event 17H , Umask 01H
343Counts the number of instructions written into the instruction queue every
344cycle.
345.It Li INST_DECODED.DEC0
346.Pq Event 18H , Umask 01H
347Counts number of instructions that require decoder 0 to be decoded. Usually,
348this means that the instruction maps to more than 1 uop
349.It Li TWO_UOP_INSTS_DECODED
350.Pq Event 19H , Umask 01H
351An instruction that generates two uops was decoded
352.It Li INST_QUEUE_WRITE_CYCLES
353.Pq Event 1EH , Umask 01H
354This event counts the number of cycles during which instructions are written
355to the instruction queue. Dividing this counter by the number of
356instructions written to the instruction queue (INST_QUEUE_WRITES) yields the
357average number of instructions decoded each cycle. If this number is less
358than four and the pipe stalls, this indicates that the decoder is failing to
359decode enough instructions per cycle to sustain the 4-wide pipeline.
360If SSE* instructions that are 6 bytes or longer arrive one after another,
361then front end throughput may limit execution speed. In such case,
362.It Li LSD_OVERFLOW
363.Pq Event 20H , Umask 01H
364Number of loops that can not stream from the instruction queue.
365.It Li L2_RQSTS.LD_HIT
366.Pq Event 24H , Umask 01H
367Counts number of loads that hit the L2 cache. L2 loads include both L1D
368demand misses as well as L1D prefetches. L2 loads can be rejected for
369various reasons. Only non rejected loads are counted.
370.It Li L2_RQSTS.LD_MISS
371.Pq Event 24H , Umask 02H
372Counts the number of loads that miss the L2 cache. L2 loads include both L1D
373demand misses as well as L1D prefetches.
374.It Li L2_RQSTS.LOADS
375.Pq Event 24H , Umask 03H
376Counts all L2 load requests. L2 loads include both L1D demand misses as well
377as L1D prefetches.
378.It Li L2_RQSTS.RFO_HIT
379.Pq Event 24H , Umask 04H
380Counts the number of store RFO requests that hit the L2 cache. L2 RFO
381requests include both L1D demand RFO misses as well as L1D RFO prefetches.
382Count includes WC memory requests, where the data is not fetched but the
383permission to write the line is required.
384.It Li L2_RQSTS.RFO_MISS
385.Pq Event 24H , Umask 08H
386Counts the number of store RFO requests that miss the L2 cache. L2 RFO
387requests include both L1D demand RFO misses as well as L1D RFO prefetches.
388.It Li L2_RQSTS.RFOS
389.Pq Event 24H , Umask 0CH
390Counts all L2 store RFO requests. L2 RFO requests include both L1D demand
391RFO misses as well as L1D RFO prefetches..
392.It Li L2_RQSTS.IFETCH_HIT
393.Pq Event 24H , Umask 10H
394Counts number of instruction fetches that hit the L2 cache. L2 instruction
395fetches include both L1I demand misses as well as L1I instruction
396prefetches.
397.It Li L2_RQSTS.IFETCH_MISS
398.Pq Event 24H , Umask 20H
399Counts number of instruction fetches that miss the L2 cache. L2 instruction
400fetches include both L1I demand misses as well as L1I instruction
401prefetches.
402.It Li L2_RQSTS.IFETCHES
403.Pq Event 24H , Umask 30H
404Counts all instruction fetches. L2 instruction fetches include both L1I
405demand misses as well as L1I instruction prefetches.
406.It Li L2_RQSTS.PREFETCH_HIT
407.Pq Event 24H , Umask 40H
408Counts L2 prefetch hits for both code and data.
409.It Li L2_RQSTS.PREFETCH_MISS
410.Pq Event 24H , Umask 80H
411Counts L2 prefetch misses for both code and data.
412.It Li L2_RQSTS.PREFETCHES
413.Pq Event 24H , Umask C0H
414Counts all L2 prefetches for both code and data.
415.It Li L2_RQSTS.MISS
416.Pq Event 24H , Umask AAH
417Counts all L2 misses for both code and data.
418.It Li L2_RQSTS.REFERENCES
419.Pq Event 24H , Umask FFH
420Counts all L2 requests for both code and data.
421.It Li L2_DATA_RQSTS.DEMAND.I_STATE
422.Pq Event 26H , Umask 01H
423Counts number of L2 data demand loads where the cache line to be loaded is
424in the I (invalid) state, i.e. a cache miss. L2 demand loads are both L1D
425demand misses and L1D prefetches.
426.It Li L2_DATA_RQSTS.DEMAND.S_STATE
427.Pq Event 26H , Umask 02H
428Counts number of L2 data demand loads where the cache line to be loaded is
429in the S (shared) state. L2 demand loads are both L1D demand misses and L1D
430prefetches.
431.It Li L2_DATA_RQSTS.DEMAND.E_STATE
432.Pq Event 26H , Umask 04H
433Counts number of L2 data demand loads where the cache line to be loaded is
434in the E (exclusive) state. L2 demand loads are both L1D demand misses and
435L1D prefetches.
436.It Li L2_DATA_RQSTS.DEMAND.M_STATE
437.Pq Event 26H , Umask 08H
438Counts number of L2 data demand loads where the cache line to be loaded is
439in the M (modified) state. L2 demand loads are both L1D demand misses and
440L1D prefetches.
441.It Li L2_DATA_RQSTS.DEMAND.MESI
442.Pq Event 26H , Umask 0FH
443Counts all L2 data demand requests. L2 demand loads are both L1D demand
444misses and L1D prefetches.
445.It Li L2_DATA_RQSTS.PREFETCH.I_STATE
446.Pq Event 26H , Umask 10H
447Counts number of L2 prefetch data loads where the cache line to be loaded is
448in the I (invalid) state, i.e. a cache miss.
449.It Li L2_DATA_RQSTS.PREFETCH.S_STATE
450.Pq Event 26H , Umask 20H
451Counts number of L2 prefetch data loads where the cache line to be loaded is
452in the S (shared) state. A prefetch RFO will miss on an S state line, while
453a prefetch read will hit on an S state line.
454.It Li L2_DATA_RQSTS.PREFETCH.E_STATE
455.Pq Event 26H , Umask 40H
456Counts number of L2 prefetch data loads where the cache line to be loaded is
457in the E (exclusive) state.
458.It Li L2_DATA_RQSTS.PREFETCH.M_STATE
459.Pq Event 26H , Umask 80H
460Counts number of L2 prefetch data loads where the cache line to be loaded is
461in the M (modified) state.
462.It Li L2_DATA_RQSTS.PREFETCH.MESI
463.Pq Event 26H , Umask F0H
464Counts all L2 prefetch requests.
465.It Li L2_DATA_RQSTS.ANY
466.Pq Event 26H , Umask FFH
467Counts all L2 data requests.
468.It Li L2_WRITE.RFO.I_STATE
469.Pq Event 27H , Umask 01H
470Counts number of L2 demand store RFO requests where the cache line to be
471loaded is in the I (invalid) state, i.e, a cache miss. The L1D prefetcher
472does not issue a RFO prefetch.
473This is a demand RFO request
474.It Li L2_WRITE.RFO.S_STATE
475.Pq Event 27H , Umask 02H
476Counts number of L2 store RFO requests where the cache line to be loaded is
477in the S (shared) state. The L1D prefetcher does not issue a RFO prefetch,.
478This is a demand RFO request
479.It Li L2_WRITE.RFO.M_STATE
480.Pq Event 27H , Umask 08H
481Counts number of L2 store RFO requests where the cache line to be loaded is
482in the M (modified) state. The L1D prefetcher does not issue a RFO prefetch.
483This is a demand RFO request
484.It Li L2_WRITE.RFO.HIT
485.Pq Event 27H , Umask 0EH
486Counts number of L2 store RFO requests where the cache line to be loaded is
487in either the S, E or M states. The L1D prefetcher does not issue a RFO
488prefetch.
489This is a demand RFO request
490.It Li L2_WRITE.RFO.MESI
491.Pq Event 27H , Umask 0FH
492Counts all L2 store RFO requests.The L1D prefetcher does not issue a RFO
493prefetch.
494This is a demand RFO request
495.It Li L2_WRITE.LOCK.I_STATE
496.Pq Event 27H , Umask 10H
497Counts number of L2 demand lock RFO requests where the cache line to be
498loaded is in the I (invalid) state, i.e. a cache miss.
499.It Li L2_WRITE.LOCK.S_STATE
500.Pq Event 27H , Umask 20H
501Counts number of L2 lock RFO requests where the cache line to be loaded is
502in the S (shared) state.
503.It Li L2_WRITE.LOCK.E_STATE
504.Pq Event 27H , Umask 40H
505Counts number of L2 demand lock RFO requests where the cache line to be
506loaded is in the E (exclusive) state.
507.It Li L2_WRITE.LOCK.M_STATE
508.Pq Event 27H , Umask 80H
509Counts number of L2 demand lock RFO requests where the cache line to be
510loaded is in the M (modified) state.
511.It Li L2_WRITE.LOCK.HIT
512.Pq Event 27H , Umask E0H
513Counts number of L2 demand lock RFO requests where the cache line to be
514loaded is in either the S, E, or M state.
515.It Li L2_WRITE.LOCK.MESI
516.Pq Event 27H , Umask F0H
517Counts all L2 demand lock RFO requests.
518.It Li L1D_WB_L2.I_STATE
519.Pq Event 28H , Umask 01H
520Counts number of L1 writebacks to the L2 where the cache line to be written
521is in the I (invalid) state, i.e. a cache miss.
522.It Li L1D_WB_L2.S_STATE
523.Pq Event 28H , Umask 02H
524Counts number of L1 writebacks to the L2 where the cache line to be written
525is in the S state.
526.It Li L1D_WB_L2.E_STATE
527.Pq Event 28H , Umask 04H
528Counts number of L1 writebacks to the L2 where the cache line to be written
529is in the E (exclusive) state.
530.It Li L1D_WB_L2.M_STATE
531.Pq Event 28H , Umask 08H
532Counts number of L1 writebacks to the L2 where the cache line to be written
533is in the M (modified) state.
534.It Li L1D_WB_L2.MESI
535.Pq Event 28H , Umask 0FH
536Counts all L1 writebacks to the L2.
537.It Li L3_LAT_CACHE.REFERENCE
538.Pq Event 2EH , Umask 02H
539Counts uncore Last Level Cache references. Because cache hierarchy, cache
540sizes and other implementation-specific characteristics; value comparison to
541estimate performance differences is not recommended.
542see Table A-1
543.It Li L3_LAT_CACHE.MISS
544.Pq Event 2EH , Umask 01H
545Counts uncore Last Level Cache misses. Because cache hierarchy, cache sizes
546and other implementation-specific characteristics; value comparison to
547estimate performance differences is not recommended.
548see Table A-1
549.It Li CPU_CLK_UNHALTED.THREAD_P
550.Pq Event 3CH , Umask 00H
551Counts the number of thread cycles while the thread is not in a halt state.
552The thread enters the halt state when it is running the HLT instruction. The
553core frequency may change from time to time due to power or thermal
554throttling.
555see Table A-1
556.It Li CPU_CLK_UNHALTED.REF_P
557.Pq Event 3CH , Umask 01H
558Increments at the frequency of TSC when not halted.
559see Table A-1
560.It Li DTLB_MISSES.ANY
561.Pq Event 49H , Umask 01H
562Counts the number of misses in the STLB which causes a page walk.
563.It Li DTLB_MISSES.WALK_COMPLETED
564.Pq Event 49H , Umask 02H
565Counts number of misses in the STLB which resulted in a completed page walk.
566.It Li DTLB_MISSES.WALK_CYCLES
567.Pq Event 49H , Umask 04H
568Counts cycles of page walk due to misses in the STLB.
569.It Li DTLB_MISSES.STLB_HIT
570.Pq Event 49H , Umask 10H
571Counts the number of DTLB first level misses that hit in the second level
572TLB. This event is only relevant if the core contains multiple DTLB levels.
573.It Li DTLB_MISSES.LARGE_WALK_COMPLETED
574.Pq Event 49H , Umask 80H
575Counts number of completed large page walks due to misses in the STLB.
576.It Li LOAD_HIT_PRE
577.Pq Event 4CH , Umask 01H
578Counts load operations sent to the L1 data cache while a previous SSE
579prefetch instruction to the same cache line has started prefetching but has
580not yet finished.
581.It Li L1D_PREFETCH.REQUESTS
582.Pq Event 4EH , Umask 01H
583Counts number of hardware prefetch requests dispatched out of the prefetch
584FIFO.
585.It Li L1D_PREFETCH.MISS
586.Pq Event 4EH , Umask 02H
587Counts number of hardware prefetch requests that miss the L1D. There are two
588prefetchers in the L1D. A streamer, which predicts lines sequentially after
589this one should be fetched, and the IP prefetcher that remembers access
590patterns for the current instruction. The streamer prefetcher stops on an
591L1D hit, while the IP prefetcher does not.
592.It Li L1D_PREFETCH.TRIGGERS
593.Pq Event 4EH , Umask 04H
594Counts number of prefetch requests triggered by the Finite State Machine and
595pushed into the prefetch FIFO. Some of the prefetch requests are dropped due
596to overwrites or competition between the IP index prefetcher and streamer
597prefetcher. The prefetch FIFO contains 4 entries.
598.It Li EPT.WALK_CYCLES
599.Pq Event 4FH , Umask 10H
600Counts Extended Page walk cycles.
601.It Li L1D.REPL
602.Pq Event 51H , Umask 01H
603Counts the number of lines brought into the L1 data cache.
604Counter 0, 1 only
605.It Li L1D.M_REPL
606.Pq Event 51H , Umask 02H
607Counts the number of modified lines brought into the L1 data cache.
608Counter 0, 1 only
609.It Li L1D.M_EVICT
610.Pq Event 51H , Umask 04H
611Counts the number of modified lines evicted from the L1 data cache due to
612replacement.
613Counter 0, 1 only
614.It Li L1D.M_SNOOP_EVICT
615.Pq Event 51H , Umask 08H
616Counts the number of modified lines evicted from the L1 data cache due to
617snoop HITM intervention.
618Counter 0, 1 only
619.It Li L1D_CACHE_PREFETCH_LOCK_FB_HIT
620.Pq Event 52H , Umask 01H
621Counts the number of cacheable load lock speculated instructions accepted
622into the fill buffer.
623.It Li L1D_CACHE_LOCK_FB_HIT
624.Pq Event 53H , Umask 01H
625Counts the number of cacheable load lock speculated or retired instructions
626accepted into the fill buffer.
627.It Li OFFCORE_REQUESTS_OUTSTANDING.DEMAND.READ_DATA
628.Pq Event 60H , Umask 01H
629Counts weighted cycles of offcore demand data read requests. Does not
630include L2 prefetch requests.
631counter 0
632.It Li OFFCORE_REQUESTS_OUTSTANDING.DEMAND.READ_CODE
633.Pq Event 60H , Umask 02H
634Counts weighted cycles of offcore demand code read requests. Does not
635include L2 prefetch requests.
636counter 0
637.It Li OFFCORE_REQUESTS_OUTSTANDING.DEMAND.RFO
638.Pq Event 60H , Umask 04H
639Counts weighted cycles of offcore demand RFO requests. Does not include L2
640prefetch requests.
641counter 0
642.It Li OFFCORE_REQUESTS_OUTSTANDING.ANY.READ
643.Pq Event 60H , Umask 08H
644Counts weighted cycles of offcore read requests of any kind. Include L2
645prefetch requests.
646counter 0
647.It Li CACHE_LOCK_CYCLES.L1D_L2
648.Pq Event 63H , Umask 01H
649Cycle count during which the L1D and L2 are locked. A lock is asserted when
650there is a locked memory access, due to uncacheable memory, a locked
651operation that spans two cache lines, or a page walk from an uncacheable
652page table.
653Counter 0, 1 only. L1D and L2 locks have a very high performance penalty and
654it is highly recommended to avoid such accesses.
655.It Li CACHE_LOCK_CYCLES.L1D
656.Pq Event 63H , Umask 02H
657Counts the number of cycles that cacheline in the L1 data cache unit is
658locked.
659Counter 0, 1 only.
660.It Li IO_TRANSACTIONS
661.Pq Event 6CH , Umask 01H
662Counts the number of completed I/O transactions.
663.It Li L1I.HITS
664.Pq Event 80H , Umask 01H
665Counts all instruction fetches that hit the L1 instruction cache.
666.It Li L1I.MISSES
667.Pq Event 80H , Umask 02H
668Counts all instruction fetches that miss the L1I cache. This includes
669instruction cache misses, streaming buffer misses, victim cache misses and
670uncacheable fetches. An instruction fetch miss is counted only once and not
671once for every cycle it is outstanding.
672.It Li L1I.READS
673.Pq Event 80H , Umask 03H
674Counts all instruction fetches, including uncacheable fetches that bypass
675the L1I.
676.It Li L1I.CYCLES_STALLED
677.Pq Event 80H , Umask 04H
678Cycle counts for which an instruction fetch stalls due to a L1I cache miss,
679ITLB miss or ITLB fault.
680.It Li LARGE_ITLB.HIT
681.Pq Event 82H , Umask 01H
682Counts number of large ITLB hits.
683.It Li ITLB_MISSES.ANY
684.Pq Event 85H , Umask 01H
685Counts the number of misses in all levels of the ITLB which causes a page
686walk.
687.It Li ITLB_MISSES.WALK_COMPLETED
688.Pq Event 85H , Umask 02H
689Counts number of misses in all levels of the ITLB which resulted in a
690completed page walk.
691.It Li ITLB_MISSES.WALK_CYCLES
692.Pq Event 85H , Umask 04H
693Counts ITLB miss page walk cycles.
694.It Li ITLB_MISSES.LARGE_WALK_COMPLETED
695.Pq Event 85H , Umask 80H
696Counts number of completed large page walks due to misses in the STLB.
697.It Li ILD_STALL.LCP
698.Pq Event 87H , Umask 01H
699Cycles Instruction Length Decoder stalls due to length changing prefixes:
70066, 67 or REX.W (for EM64T) instructions which change the length of the
701decoded instruction.
702.It Li ILD_STALL.MRU
703.Pq Event 87H , Umask 02H
704Instruction Length Decoder stall cycles due to Brand Prediction Unit (PBU)
705Most Recently Used (MRU) bypass.
706.It Li ILD_STALL.IQ_FULL
707.Pq Event 87H , Umask 04H
708Stall cycles due to a full instruction queue.
709.It Li ILD_STALL.REGEN
710.Pq Event 87H , Umask 08H
711Counts the number of regen stalls.
712.It Li ILD_STALL.ANY
713.Pq Event 87H , Umask 0FH
714Counts any cycles the Instruction Length Decoder is stalled.
715.It Li BR_INST_EXEC.COND
716.Pq Event 88H , Umask 01H
717Counts the number of conditional near branch instructions executed, but not
718necessarily retired.
719.It Li BR_INST_EXEC.DIRECT
720.Pq Event 88H , Umask 02H
721Counts all unconditional near branch instructions excluding calls and
722indirect branches.
723.It Li BR_INST_EXEC.INDIRECT_NON_CALL
724.Pq Event 88H , Umask 04H
725Counts the number of executed indirect near branch instructions that are not
726calls.
727.It Li BR_INST_EXEC.NON_CALLS
728.Pq Event 88H , Umask 07H
729Counts all non call near branch instructions executed, but not necessarily
730retired.
731.It Li BR_INST_EXEC.RETURN_NEAR
732.Pq Event 88H , Umask 08H
733Counts indirect near branches that have a return mnemonic.
734.It Li BR_INST_EXEC.DIRECT_NEAR_CALL
735.Pq Event 88H , Umask 10H
736Counts unconditional near call branch instructions, excluding non call
737branch, executed.
738.It Li BR_INST_EXEC.INDIRECT_NEAR_CALL
739.Pq Event 88H , Umask 20H
740Counts indirect near calls, including both register and memory indirect,
741executed.
742.It Li BR_INST_EXEC.NEAR_CALLS
743.Pq Event 88H , Umask 30H
744Counts all near call branches executed, but not necessarily retired.
745.It Li BR_INST_EXEC.TAKEN
746.Pq Event 88H , Umask 40H
747Counts taken near branches executed, but not necessarily retired.
748.It Li BR_INST_EXEC.ANY
749.Pq Event 88H , Umask 7FH
750Counts all near executed branches (not necessarily retired). This includes
751only instructions and not micro-op branches. Frequent branching is not
752necessarily a major performance issue. However frequent branch
753mispredictions may be a problem.
754.It Li BR_MISP_EXEC.COND
755.Pq Event 89H , Umask 01H
756Counts the number of mispredicted conditional near branch instructions
757executed, but not necessarily retired.
758.It Li BR_MISP_EXEC.DIRECT
759.Pq Event 89H , Umask 02H
760Counts mispredicted macro unconditional near branch instructions, excluding
761calls and indirect branches (should always be 0).
762.It Li BR_MISP_EXEC.INDIRECT_NON_CALL
763.Pq Event 89H , Umask 04H
764Counts the number of executed mispredicted indirect near branch instructions
765that are not calls.
766.It Li BR_MISP_EXEC.NON_CALLS
767.Pq Event 89H , Umask 07H
768Counts mispredicted non call near branches executed, but not necessarily
769retired.
770.It Li BR_MISP_EXEC.RETURN_NEAR
771.Pq Event 89H , Umask 08H
772Counts mispredicted indirect branches that have a rear return mnemonic.
773.It Li BR_MISP_EXEC.DIRECT_NEAR_CALL
774.Pq Event 89H , Umask 10H
775Counts mispredicted non-indirect near calls executed, (should always be 0).
776.It Li BR_MISP_EXEC.INDIRECT_NEAR_CALL
777.Pq Event 89H , Umask 20H
778Counts mispredicted indirect near calls executed, including both register
779and memory indirect.
780.It Li BR_MISP_EXEC.NEAR_CALLS
781.Pq Event 89H , Umask 30H
782Counts all mispredicted near call branches executed, but not necessarily
783retired.
784.It Li BR_MISP_EXEC.TAKEN
785.Pq Event 89H , Umask 40H
786Counts executed mispredicted near branches that are taken, but not
787necessarily retired.
788.It Li BR_MISP_EXEC.ANY
789.Pq Event 89H , Umask 7FH
790Counts the number of mispredicted near branch instructions that were
791executed, but not necessarily retired.
792.It Li RESOURCE_STALLS.ANY
793.Pq Event A2H , Umask 01H
794Counts the number of Allocator resource related stalls. Includes register
795renaming buffer entries, memory buffer entries. In addition to resource
796related stalls, this event counts some other events. Includes stalls arising
797during branch misprediction recovery, such as if retirement of the
798mispredicted branch is delayed and stalls arising while store buffer is
799draining from synchronizing operations.
800Does not include stalls due to SuperQ (off core) queue full, too many cache
801misses, etc.
802.It Li RESOURCE_STALLS.LOAD
803.Pq Event A2H , Umask 02H
804Counts the cycles of stall due to lack of load buffer for load operation.
805.It Li RESOURCE_STALLS.RS_FULL
806.Pq Event A2H , Umask 04H
807This event counts the number of cycles when the number of instructions in
808the pipeline waiting for execution reaches the limit the processor can
809handle. A high count of this event indicates that there are long latency
810operations in the pipe (possibly load and store operations that miss the L2
811cache, or instructions dependent upon instructions further down the pipeline
812that have yet to retire.
813When RS is full, new instructions can not enter the reservation station and
814start execution.
815.It Li RESOURCE_STALLS.STORE
816.Pq Event A2H , Umask 08H
817This event counts the number of cycles that a resource related stall will
818occur due to the number of store instructions reaching the limit of the
819pipeline, (i.e. all store buffers are used). The stall ends when a store
820instruction commits its data to the cache or memory.
821.It Li RESOURCE_STALLS.ROB_FULL
822.Pq Event A2H , Umask 10H
823Counts the cycles of stall due to re- order buffer full.
824.It Li RESOURCE_STALLS.FPCW
825.Pq Event A2H , Umask 20H
826Counts the number of cycles while execution was stalled due to writing the
827floating-point unit (FPU) control word.
828.It Li RESOURCE_STALLS.MXCSR
829.Pq Event A2H , Umask 40H
830Stalls due to the MXCSR register rename occurring to close to a previous
831MXCSR rename. The MXCSR provides control and status for the MMX registers.
832.It Li RESOURCE_STALLS.OTHER
833.Pq Event A2H , Umask 80H
834Counts the number of cycles while execution was stalled due to other
835resource issues.
836.It Li MACRO_INSTS.FUSIONS_DECODED
837.Pq Event A6H , Umask 01H
838Counts the number of instructions decoded that are macro-fused but not
839necessarily executed or retired.
840.It Li BACLEAR_FORCE_IQ
841.Pq Event A7H , Umask 01H
842Counts number of times a BACLEAR was forced by the Instruction Queue. The IQ
843is also responsible for providing conditional branch prediction direction
844based on a static scheme and dynamic data provided by the L2 Branch
845Prediction Unit. If the conditional branch target is not found in the Target
846Array and the IQ predicts that the branch is taken, then the IQ will force
847the Branch Address Calculator to issue a BACLEAR. Each BACLEAR asserted by
848the BAC generates approximately an 8 cycle bubble in the instruction fetch
849pipeline.
850.It Li LSD.UOPS
851.Pq Event A8H , Umask 01H
852Counts the number of micro-ops delivered by loop stream detector
853Use cmask=1 and invert to count cycles
854.It Li ITLB_FLUSH
855.Pq Event AEH , Umask 01H
856Counts the number of ITLB flushes
857.It Li OFFCORE_REQUESTS.DEMAND.READ_DATA
858.Pq Event B0H , Umask 01H
859Counts number of offcore demand data read requests. Does not count L2
860prefetch requests.
861.It Li OFFCORE_REQUESTS.DEMAND.READ_CODE
862.Pq Event B0H , Umask 02H
863Counts number of offcore demand code read requests. Does not count L2
864prefetch requests.
865.It Li OFFCORE_REQUESTS.DEMAND.RFO
866.Pq Event B0H , Umask 04H
867Counts number of offcore demand RFO requests. Does not count L2 prefetch
868requests.
869.It Li OFFCORE_REQUESTS.ANY.READ
870.Pq Event B0H , Umask 08H
871Counts number of offcore read requests. Includes L2 prefetch requests.
872.It Li OFFCORE_REQUESTS.ANY.RFO
873.Pq Event 80H , Umask 10H
874Counts number of offcore RFO requests. Includes L2 prefetch requests.
875.It Li OFFCORE_REQUESTS.L1D_WRITEBACK
876.Pq Event B0H , Umask 40H
877Counts number of L1D writebacks to the uncore.
878.It Li OFFCORE_REQUESTS.ANY
879.Pq Event B0H , Umask 80H
880Counts all offcore requests.
881.It Li UOPS_EXECUTED.PORT0
882.Pq Event B1H , Umask 01H
883Counts number of Uops executed that were issued on port 0. Port 0 handles
884integer arithmetic, SIMD and FP add Uops.
885.It Li UOPS_EXECUTED.PORT1
886.Pq Event B1H , Umask 02H
887Counts number of Uops executed that were issued on port 1. Port 1 handles
888integer arithmetic, SIMD, integer shift, FP multiply and FP divide Uops.
889.It Li UOPS_EXECUTED.PORT2_CORE
890.Pq Event B1H , Umask 04H
891Counts number of Uops executed that were issued on port 2. Port 2 handles
892the load Uops. This is a core count only and can not be collected per
893thread.
894.It Li UOPS_EXECUTED.PORT3_CORE
895.Pq Event B1H , Umask 08H
896Counts number of Uops executed that were issued on port 3. Port 3 handles
897store Uops. This is a core count only and can not be collected per thread.
898.It Li UOPS_EXECUTED.PORT4_CORE
899.Pq Event B1H , Umask 10H
900Counts number of Uops executed that where issued on port 4. Port 4 handles
901the value to be stored for the store Uops issued on port 3. This is a core
902count only and can not be collected per thread.
903.It Li UOPS_EXECUTED.CORE_ACTIVE_CYCLES_NO_PORT5
904.Pq Event B1H , Umask 1FH
905Counts number of cycles there are one or more uops being executed and were
906issued on ports 0-4. This is a core count only and can not be collected per
907thread.
908.It Li UOPS_EXECUTED.PORT5
909.Pq Event B1H , Umask 20H
910Counts number of Uops executed that where issued on port 5.
911.It Li UOPS_EXECUTED.CORE_ACTIVE_CYCLES
912.Pq Event B1H , Umask 3FH
913Counts number of cycles there are one or more uops being executed on any
914ports. This is a core count only and can not be collected per thread.
915.It Li UOPS_EXECUTED.PORT015
916.Pq Event B1H , Umask 40H
917Counts number of Uops executed that where issued on port 0, 1, or 5.
918use cmask=1, invert=1 to count stall cycles
919.It Li UOPS_EXECUTED.PORT234
920.Pq Event B1H , Umask 80H
921Counts number of Uops executed that where issued on port 2, 3, or 4.
922.It Li OFFCORE_REQUESTS_SQ_FULL
923.Pq Event B2H , Umask 01H
924Counts number of cycles the SQ is full to handle off-core requests.
925.It Li SNOOPQ_REQUESTS_OUTSTANDING.DATA
926.Pq Event B3H , Umask 01H
927Counts weighted cycles of snoopq requests for data. Counter 0 only
928Use cmask=1 to count cycles not empty.
929.It Li SNOOPQ_REQUESTS_OUTSTANDING.INVALIDATE
930.Pq Event B3H , Umask 02H
931Counts weighted cycles of snoopq invalidate requests. Counter 0 only
932Use cmask=1 to count cycles not empty.
933.It Li SNOOPQ_REQUESTS_OUTSTANDING.CODE
934.Pq Event B3H , Umask 04H
935Counts weighted cycles of snoopq requests for code. Counter 0 only
936Use cmask=1 to count cycles not empty.
937.It Li SNOOPQ_REQUESTS.CODE
938.Pq Event B4H , Umask 01H
939Counts the number of snoop code requests
940.It Li SNOOPQ_REQUESTS.DATA
941.Pq Event B4H , Umask 02H
942Counts the number of snoop data requests
943.It Li SNOOPQ_REQUESTS.INVALIDATE
944.Pq Event B4H , Umask 04H
945Counts the number of snoop invalidate requests
946.It Li OFF_CORE_RESPONSE_0
947.Pq Event B7H , Umask 01H
948see Section 30.6.1.3, Off-core Response Performance Monitoring in the
949Processor Core.
950Requires programming MSR 01A6H
951.It Li SNOOP_RESPONSE.HIT
952.Pq Event B8H , Umask 01H
953Counts HIT snoop response sent by this thread in response to a snoop
954request.
955.It Li SNOOP_RESPONSE.HITE
956.Pq Event B8H , Umask 02H
957Counts HIT E snoop response sent by this thread in response to a snoop
958request.
959.It Li SNOOP_RESPONSE.HITM
960.Pq Event B8H , Umask 04H
961Counts HIT M snoop response sent by this thread in response to a snoop
962request.
963.It Li OFF_CORE_RESPONSE_1
964.Pq Event BBH , Umask 01H
965see Section 30.6.1.3, Off-core Response Performance Monitoring in the
966Processor Core
967Use MSR 01A7H
968.It Li INST_RETIRED.ANY_P
969.Pq Event C0H , Umask 01H
970See Table A-1
971Notes: INST_RETIRED.ANY is counted by a designated fixed counter.
972INST_RETIRED.ANY_P is counted by a programmable counter and is an
973architectural performance event. Event is supported if CPUID.A.EBX[1] = 0.
974Counting: Faulting executions of GETSEC/VM entry/VM Exit/MWait will not
975count as retired instructions.
976.It Li INST_RETIRED.X87
977.Pq Event C0H , Umask 02H
978Counts the number of floating point computational operations retired
979floating point computational operations executed by the assist handler and
980sub-operations of complex floating point instructions like transcendental
981instructions.
982.It Li INST_RETIRED.MMX
983.Pq Event C0H , Umask 04H
984Counts the number of retired: MMX instructions.
985.It Li UOPS_RETIRED.ANY
986.Pq Event C2H , Umask 01H
987Counts the number of micro-ops retired, (macro-fused=1, micro- fused=2,
988others=1; maximum count of 8 per cycle). Most instructions are composed of
989one or two micro-ops. Some instructions are decoded into longer sequences
990such as repeat instructions, floating point transcendental instructions, and
991assists.
992Use cmask=1 and invert to count active cycles or stalled cycles
993.It Li UOPS_RETIRED.RETIRE_SLOTS
994.Pq Event C2H , Umask 02H
995Counts the number of retirement slots used each cycle
996.It Li UOPS_RETIRED.MACRO_FUSED
997.Pq Event C2H , Umask 04H
998Counts number of macro-fused uops retired.
999.It Li MACHINE_CLEARS.CYCLES
1000.Pq Event C3H , Umask 01H
1001Counts the cycles machine clear is asserted.
1002.It Li MACHINE_CLEARS.MEM_ORDER
1003.Pq Event C3H , Umask 02H
1004Counts the number of machine clears due to memory order conflicts.
1005.It Li MACHINE_CLEARS.SMC
1006.Pq Event C3H , Umask 04H
1007Counts the number of times that a program writes to a code section.
1008Self-modifying code causes a sever penalty in all Intel 64 and IA-32
1009processors. The modified cache line is written back to the L2 and L3caches.
1010.It Li BR_INST_RETIRED.ALL_BRANCHES
1011.Pq Event C4H , Umask 00H
1012See Table A-1
1013.It Li BR_INST_RETIRED.CONDITIONAL
1014.Pq Event C4H , Umask 01H
1015Counts the number of conditional branch instructions retired.
1016.It Li BR_INST_RETIRED.NEAR_CALL
1017.Pq Event C4H , Umask 02H
1018Counts the number of direct & indirect near unconditional calls retired
1019.It Li BR_INST_RETIRED.ALL_BRANCHES
1020.Pq Event C4H , Umask 04H
1021Counts the number of branch instructions retired
1022.It Li BR_MISP_RETIRED.ALL_BRANCHES
1023.Pq Event C5H , Umask 00H
1024See Table A-1
1025.It Li BR_MISP_RETIRED.CONDITIONAL
1026.Pq Event C5H , Umask 01H
1027Counts mispredicted conditional retired calls.
1028.It Li BR_MISP_RETIRED.NEAR_CALL
1029.Pq Event C5H , Umask 02H
1030Counts mispredicted direct & indirect near unconditional retired calls.
1031.It Li BR_MISP_RETIRED.ALL_BRANCHES
1032.Pq Event C5H , Umask 04H
1033Counts all mispredicted retired calls.
1034.It Li SSEX_UOPS_RETIRED.PACKED_SINGLE
1035.Pq Event C7H , Umask 01H
1036Counts SIMD packed single-precision floating point Uops retired.
1037.It Li SSEX_UOPS_RETIRED.SCALAR_SINGLE
1038.Pq Event C7H , Umask 02H
1039Counts SIMD calar single-precision floating point Uops retired.
1040.It Li SSEX_UOPS_RETIRED.PACKED_DOUBLE
1041.Pq Event C7H , Umask 04H
1042Counts SIMD packed double- precision floating point Uops retired.
1043.It Li SSEX_UOPS_RETIRED.SCALAR_DOUBLE
1044.Pq Event C7H , Umask 08H
1045Counts SIMD scalar double-precision floating point Uops retired.
1046.It Li SSEX_UOPS_RETIRED.VECTOR_INTEGER
1047.Pq Event C7H , Umask 10H
1048Counts 128-bit SIMD vector integer Uops retired.
1049.It Li ITLB_MISS_RETIRED
1050.Pq Event C8H , Umask 20H
1051Counts the number of retired instructions that missed the ITLB when the
1052instruction was fetched.
1053.It Li MEM_LOAD_RETIRED.L1D_HIT
1054.Pq Event CBH , Umask 01H
1055Counts number of retired loads that hit the L1 data cache.
1056.It Li MEM_LOAD_RETIRED.L2_HIT
1057.Pq Event CBH , Umask 02H
1058Counts number of retired loads that hit the L2 data cache.
1059.It Li MEM_LOAD_RETIRED.L3_UNSHARED_HIT
1060.Pq Event CBH , Umask 04H
1061Counts number of retired loads that hit their own, unshared lines in the L3
1062cache.
1063.It Li MEM_LOAD_RETIRED.OTHER_CORE_L2_HIT_HITM
1064.Pq Event CBH , Umask 08H
1065Counts number of retired loads that hit in a sibling core's L2 (on die
1066core). Since the L3 is inclusive of all cores on the package, this is an L3
1067hit. This counts both clean or modified hits.
1068.It Li MEM_LOAD_RETIRED.L3_MISS
1069.Pq Event CBH , Umask 10H
1070Counts number of retired loads that miss the L3 cache. The load was
1071satisfied by a remote socket, local memory or an IOH.
1072.It Li MEM_LOAD_RETIRED.HIT_LFB
1073.Pq Event CBH , Umask 40H
1074Counts number of retired loads that miss the L1D and the address is located
1075in an allocated line fill buffer and will soon be committed to cache. This
1076is counting secondary L1D misses.
1077.It Li MEM_LOAD_RETIRED.DTLB_MISS
1078.Pq Event CBH , Umask 80H
1079Counts the number of retired loads that missed the DTLB. The DTLB miss is
1080not counted if the load operation causes a fault. This event counts loads
1081from cacheable memory only. The event does not count loads by software
1082prefetches. Counts both primary and secondary misses to the TLB.
1083.It Li FP_MMX_TRANS.TO_FP
1084.Pq Event CCH , Umask 01H
1085Counts the first floating-point instruction following any MMX instruction.
1086You can use this event to estimate the penalties for the transitions between
1087floating-point and MMX technology states.
1088.It Li FP_MMX_TRANS.TO_MMX
1089.Pq Event CCH , Umask 02H
1090Counts the first MMX instruction following a floating-point instruction. You
1091can use this event to estimate the penalties for the transitions between
1092floating-point and MMX technology states.
1093.It Li FP_MMX_TRANS.ANY
1094.Pq Event CCH , Umask 03H
1095Counts all transitions from floating point to MMX instructions and from MMX
1096instructions to floating point instructions. You can use this event to
1097estimate the penalties for the transitions between floating-point and MMX
1098technology states.
1099.It Li MACRO_INSTS.DECODED
1100.Pq Event D0H , Umask 01H
1101Counts the number of instructions decoded, (but not necessarily executed or
1102retired).
1103.It Li UOPS_DECODED.STALL_CYCLES
1104.Pq Event D1H , Umask 01H
1105Counts the cycles of decoder stalls.
1106.It Li UOPS_DECODED.MS
1107.Pq Event D1H , Umask 02H
1108Counts the number of Uops decoded by the Microcode Sequencer, MS. The MS
1109delivers uops when the instruction is more than 4 uops long or a microcode
1110assist is occurring.
1111.It Li UOPS_DECODED.ESP_FOLDING
1112.Pq Event D1H , Umask 04H
1113Counts number of stack pointer (ESP) instructions decoded: push , pop , call
1114, ret, etc. ESP instructions do not generate a Uop to increment or decrement
1115ESP. Instead, they update an ESP_Offset register that keeps track of the
1116delta to the current value of the ESP register.
1117.It Li UOPS_DECODED.ESP_SYNC
1118.Pq Event D1H , Umask 08H
1119Counts number of stack pointer (ESP) sync operations where an ESP
1120instruction is corrected by adding the ESP offset register to the current
1121value of the ESP register.
1122.It Li RAT_STALLS.FLAGS
1123.Pq Event D2H , Umask 01H
1124Counts the number of cycles during which execution stalled due to several
1125reasons, one of which is a partial flag register stall. A partial register
1126stall may occur when two conditions are met: 1) an instruction modifies
1127some, but not all, of the flags in the flag register and 2) the next
1128instruction, which depends on flags, depends on flags that were not modified
1129by this instruction.
1130.It Li RAT_STALLS.REGISTERS
1131.Pq Event D2H , Umask 02H
1132This event counts the number of cycles instruction execution latency became
1133longer than the defined latency because the instruction used a register that
1134was partially written by previous instruction.
1135.It Li RAT_STALLS.ROB_READ_PORT
1136.Pq Event D2H , Umask 04H
1137Counts the number of cycles when ROB read port stalls occurred, which did
1138not allow new micro-ops to enter the out-of-order pipeline. Note that, at
1139this stage in the pipeline, additional stalls may occur at the same cycle
1140and prevent the stalled micro-ops from entering the pipe. In such a case,
1141micro-ops retry entering the execution pipe in the next cycle and the
1142ROB-read port stall is counted again.
1143.It Li RAT_STALLS.SCOREBOARD
1144.Pq Event D2H , Umask 08H
1145Counts the cycles where we stall due to microarchitecturally required
1146serialization. Microcode scoreboarding stalls.
1147.It Li RAT_STALLS.ANY
1148.Pq Event D2H , Umask 0FH
1149Counts all Register Allocation Table stall cycles due to: Cycles when ROB
1150read port stalls occurred, which did not allow new micro-ops to enter the
1151execution pipe. Cycles when partial register stalls occurred Cycles when
1152flag stalls occurred Cycles floating-point unit (FPU) status word stalls
1153occurred. To count each of these conditions separately use the events:
1154RAT_STALLS.ROB_READ_PORT, RAT_STALLS.PARTIAL, RAT_STALLS.FLAGS, and
1155RAT_STALLS.FPSW.
1156.It Li SEG_RENAME_STALLS
1157.Pq Event D4H , Umask 01H
1158Counts the number of stall cycles due to the lack of renaming resources for
1159the ES, DS, FS, and GS segment registers. If a segment is renamed but not
1160retired and a second update to the same segment occurs, a stall occurs in
1161the front- end of the pipeline until the renamed segment retires.
1162.It Li ES_REG_RENAMES
1163.Pq Event D5H , Umask 01H
1164Counts the number of times the ES segment register is renamed.
1165.It Li UOP_UNFUSION
1166.Pq Event DBH , Umask 01H
1167Counts unfusion events due to floating point exception to a fused uop.
1168.It Li BR_INST_DECODED
1169.Pq Event E0H , Umask 01H
1170Counts the number of branch instructions decoded.
1171.It Li BPU_MISSED_CALL_RET
1172.Pq Event E5H , Umask 01H
1173Counts number of times the Branch Prediction Unit missed predicting a call
1174or return branch.
1175.It Li BACLEAR.CLEAR
1176.Pq Event E6H , Umask 01H
1177Counts the number of times the front end is resteered, mainly when the
1178Branch Prediction Unit cannot provide a correct prediction and this is
1179corrected by the Branch Address Calculator at the front end. This can occur
1180if the code has many branches such that they cannot be consumed by the BPU.
1181Each BACLEAR asserted by the BAC generates approximately an 8 cycle bubble
1182in the instruction fetch pipeline. The effect on total execution time
1183depends on the surrounding code.
1184.It Li BACLEAR.BAD_TARGET
1185.Pq Event E6H , Umask 02H
1186Counts number of Branch Address Calculator clears (BACLEAR) asserted due to
1187conditional branch instructions in which there was a target hit but the
1188direction was wrong. Each BACLEAR asserted by the BAC generates
1189approximately an 8 cycle bubble in the instruction fetch pipeline.
1190.It Li BPU_CLEARS.EARLY
1191.Pq Event E8H , Umask 01H
1192Counts early (normal) Branch Prediction Unit clears: BPU predicted a taken
1193branch after incorrectly assuming that it was not taken.
1194The BPU clear leads to 2 cycle bubble in the Front End.
1195.It Li BPU_CLEARS.LATE
1196.Pq Event E8H , Umask 02H
1197Counts late Branch Prediction Unit clears due to Most Recently Used
1198conflicts. The PBU clear leads to a 3 cycle bubble in the Front End.
1199.It Li THREAD_ACTIVE
1200.Pq Event ECH , Umask 01H
1201Counts cycles threads are active.
1202.It Li L2_TRANSACTIONS.LOAD
1203.Pq Event F0H , Umask 01H
1204Counts L2 load operations due to HW prefetch or demand loads.
1205.It Li L2_TRANSACTIONS.RFO
1206.Pq Event F0H , Umask 02H
1207Counts L2 RFO operations due to HW prefetch or demand RFOs.
1208.It Li L2_TRANSACTIONS.IFETCH
1209.Pq Event F0H , Umask 04H
1210Counts L2 instruction fetch operations due to HW prefetch or demand ifetch.
1211.It Li L2_TRANSACTIONS.PREFETCH
1212.Pq Event F0H , Umask 08H
1213Counts L2 prefetch operations.
1214.It Li L2_TRANSACTIONS.L1D_WB
1215.Pq Event F0H , Umask 10H
1216Counts L1D writeback operations to the L2.
1217.It Li L2_TRANSACTIONS.FILL
1218.Pq Event F0H , Umask 20H
1219Counts L2 cache line fill operations due to load, RFO, L1D writeback or
1220prefetch.
1221.It Li L2_TRANSACTIONS.WB
1222.Pq Event F0H , Umask 40H
1223Counts L2 writeback operations to the L3.
1224.It Li L2_TRANSACTIONS.ANY
1225.Pq Event F0H , Umask 80H
1226Counts all L2 cache operations.
1227.It Li L2_LINES_IN.S_STATE
1228.Pq Event F1H , Umask 02H
1229Counts the number of cache lines allocated in the L2 cache in the S (shared)
1230state.
1231.It Li L2_LINES_IN.E_STATE
1232.Pq Event F1H , Umask 04H
1233Counts the number of cache lines allocated in the L2 cache in the E
1234(exclusive) state.
1235.It Li L2_LINES_IN.ANY
1236.Pq Event F1H , Umask 07H
1237Counts the number of cache lines allocated in the L2 cache.
1238.It Li L2_LINES_OUT.DEMAND_CLEAN
1239.Pq Event F2H , Umask 01H
1240Counts L2 clean cache lines evicted by a demand request.
1241.It Li L2_LINES_OUT.DEMAND_DIRTY
1242.Pq Event F2H , Umask 02H
1243Counts L2 dirty (modified) cache lines evicted by a demand request.
1244.It Li L2_LINES_OUT.PREFETCH_CLEAN
1245.Pq Event F2H , Umask 04H
1246Counts L2 clean cache line evicted by a prefetch request.
1247.It Li L2_LINES_OUT.PREFETCH_DIRTY
1248.Pq Event F2H , Umask 08H
1249Counts L2 modified cache line evicted by a prefetch request.
1250.It Li L2_LINES_OUT.ANY
1251.Pq Event F2H , Umask 0FH
1252Counts all L2 cache lines evicted for any reason.
1253.It Li SQ_MISC.LRU_HINTS
1254.Pq Event F4H , Umask 04H
1255Counts number of Super Queue LRU hints sent to L3.
1256.It Li SQ_MISC.SPLIT_LOCK
1257.Pq Event F4H , Umask 10H
1258Counts the number of SQ lock splits across a cache line.
1259.It Li SQ_FULL_STALL_CYCLES
1260.Pq Event F6H , Umask 01H
1261Counts cycles the Super Queue is full. Neither of the threads on this core
1262will be able to access the uncore.
1263.It Li FP_ASSIST.ALL
1264.Pq Event F7H , Umask 01H
1265Counts the number of floating point operations executed that required
1266micro-code assist intervention. Assists are required in the following cases:
1267SSE instructions, (Denormal input when the DAZ flag is off or Underflow
1268result when the FTZ flag is off): x87 instructions, (NaN or denormal are
1269loaded to a register or used as input from memory, Division by 0 or
1270Underflow output).
1271.It Li FP_ASSIST.OUTPUT
1272.Pq Event F7H , Umask 02H
1273Counts number of floating point micro-code assist when the output value
1274(destination register) is invalid.
1275.It Li FP_ASSIST.INPUT
1276.Pq Event F7H , Umask 04H
1277Counts number of floating point micro-code assist when the input value (one
1278of the source operands to an FP instruction) is invalid.
1279.It Li SIMD_INT_64.PACKED_MPY
1280.Pq Event FDH , Umask 01H
1281Counts number of SID integer 64 bit packed multiply operations.
1282.It Li SIMD_INT_64.PACKED_SHIFT
1283.Pq Event FDH , Umask 02H
1284Counts number of SID integer 64 bit packed shift operations.
1285.It Li SIMD_INT_64.PACK
1286.Pq Event FDH , Umask 04H
1287Counts number of SID integer 64 bit pack operations.
1288.It Li SIMD_INT_64.UNPACK
1289.Pq Event FDH , Umask 08H
1290Counts number of SID integer 64 bit unpack operations.
1291.It Li SIMD_INT_64.PACKED_LOGICAL
1292.Pq Event FDH , Umask 10H
1293Counts number of SID integer 64 bit logical operations.
1294.It Li SIMD_INT_64.PACKED_ARITH
1295.Pq Event FDH , Umask 20H
1296Counts number of SID integer 64 bit arithmetic operations.
1297.It Li SIMD_INT_64.SHUFFLE_MOVE
1298.Pq Event FDH , Umask 40H
1299Counts number of SID integer 64 bit shift or move operations.
1300.El
1301.Sh SEE ALSO
1302.Xr pmc 3 ,
1303.Xr pmc.atom 3 ,
1304.Xr pmc.core 3 ,
1305.Xr pmc.iaf 3 ,
1306.Xr pmc.ucf 3 ,
1307.Xr pmc.k7 3 ,
1308.Xr pmc.k8 3 ,
1309.Xr pmc.p4 3 ,
1310.Xr pmc.p5 3 ,
1311.Xr pmc.p6 3 ,
1312.Xr pmc.corei7 3 ,
1313.Xr pmc.corei7uc 3 ,
1314.Xr pmc.westmereuc 3 ,
1315.Xr pmc.tsc 3 ,
1316.Xr pmc_cpuinfo 3 ,
1317.Xr pmclog 3 ,
1318.Xr hwpmc 4
1319.Sh HISTORY
1320The
1321.Nm pmc
1322library first appeared in
1323.Fx 6.0 .
1324.Sh AUTHORS
1325The
1326.Lb libpmc
1327library was written by
1328.An "Joseph Koshy"
1329.Aq jkoshy@FreeBSD.org .
1330