All of lore.kernel.org
 help / color / mirror / Atom feed
* Not getting PMC events on Broadwell-EX
@ 2017-08-08  4:37 H. Taheri
  2017-08-08 13:53 ` Vince Weaver
                   ` (2 more replies)
  0 siblings, 3 replies; 4+ messages in thread
From: H. Taheri @ 2017-08-08  4:37 UTC (permalink / raw)
  To: linux-perf-users

Hi experts,
I am getting 0s, or <not supported>, or <not counted>  for events that are supposed to work on this processor. For example, according to Intel documentation, L2_RQSTS.MISS and L2_RQSTS.REFERENCES are supported on this processor with event code 0x24 and unit masks 0x3f and 0xff. But when I run perf stat (also with -d -d -d to see if the built-in command options can get the cache stats), I get:

[root@prme-perf-broadwell-2 intel]# perf stat -d -d -d -e cycles -e instructions -e r24ff -e r243f sleep 10

Performance counter stats for 'sleep 10':

          295,132      cycles                                                     
          676,801      instructions              #    2.29  insn per cycle        
                0      r24ff                                                      
                0      r243f                                                      
          175,865      L1-dcache-loads                                            
           12,686      L1-dcache-load-misses     #    7.21% of all L1-dcache hits    (67.93%)
    <not counted>      LLC-loads                                                     (0.00%)
    <not counted>      LLC-load-misses                                               (0.00%)
  <not supported>      L1-icache-loads                                            
    <not counted>      L1-icache-load-misses                                         (0.00%)
    <not counted>      dTLB-loads                                                    (0.00%)
    <not counted>      dTLB-load-misses                                              (0.00%)
    <not counted>      iTLB-loads                                                    (0.00%)
    <not counted>      iTLB-load-misses                                              (0.00%)
  <not supported>      L1-dcache-prefetches                                       
  <not supported>      L1-dcache-prefetch-misses                                  

     10.001308205 seconds time elapsed


This is the /proc/cpuinfo from the guest:

processor       : 0
vendor_id       : GenuineIntel
cpu family      : 6
model           : 79
model name      : Intel(R) Xeon(R) CPU E7-8890 v4 @ 2.20GHz
stepping        : 1
microcode       : 0xb00001d
cpu MHz         : 1200.031
cache size      : 30720 KB
physical id     : 0
siblings        : 48
core id         : 0
cpu cores       : 24
apicid          : 0
initial apicid  : 0
fpu             : yes
fpu_exception   : yes
cpuid level     : 20
wp              : yes
flags           : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss ht tm pbe syscall nx pdpe1gb rdtscp lm constant_tsc arch_perfmon pebs bts rep_good nopl xtopology nonstop_tsc aperfmperf eagerfpu pni pclmulqdq dtes64 monitor ds_cpl vmx smx est tm2 ssse3 fma cx16 xtpr pdcm pcid dca sse4_1 sse4_2 x2apic movbe popcnt tsc_deadline_timer aes xsave avx f16c rdrand lahf_lm abm 3dnowprefetch ida arat epb pln pts dtherm intel_pt tpr_shadow vnmi flexpriority ept vpid fsgsbase tsc_adjust bmi1 hle avx2 smep bmi2 erms invpcid rtm cqm rdseed adx smap xsaveopt cqm_llc cqm_occup_llc cqm_mbm_total cqm_mbm_local
bogomips        : 4388.87
clflush size    : 64
cache_alignment : 64
address sizes   : 46 bits physical, 48 bits virtual
power management:

Thanks,
Reza

^ permalink raw reply	[flat|nested] 4+ messages in thread

* Re: Not getting PMC events on Broadwell-EX
  2017-08-08  4:37 Not getting PMC events on Broadwell-EX H. Taheri
@ 2017-08-08 13:53 ` Vince Weaver
  2017-08-08 16:32 ` Andi Kleen
  2017-08-08 16:54 ` Reza Taheri
  2 siblings, 0 replies; 4+ messages in thread
From: Vince Weaver @ 2017-08-08 13:53 UTC (permalink / raw)
  To: H. Taheri; +Cc: linux-perf-users

On Mon, 7 Aug 2017, H. Taheri wrote:

> I am getting 0s, or <not supported>, or <not counted> for events that 
> are supposed to work on this processor. For example, according to Intel 
> documentation, L2_RQSTS.MISS and L2_RQSTS.REFERENCES are supported on 
> this processor with event code 0x24 and unit masks 0x3f and 0xff. But 
> when I run perf stat (also with -d -d -d to see if the built-in command 
> options can get the cache stats), I get:

Check your docs on the order for the fields.

You can use libpfm4 as so:
	./check_events L2_RQSTS:MISS
	Requested Event: L2_RQSTS:MISS
	Actual    Event: 
	bdw_ep::L2_RQSTS:MISS:k=1:u=1:e=0:i=0:c=0:t=0:intx=0:intxcp=0
	PMU            : Intel Broadwell EP
	IDX            : 421527583
	Codes          : 0x533f24

You want the umask first, then the event (0x3f24)

Vince

^ permalink raw reply	[flat|nested] 4+ messages in thread

* Re: Not getting PMC events on Broadwell-EX
  2017-08-08  4:37 Not getting PMC events on Broadwell-EX H. Taheri
  2017-08-08 13:53 ` Vince Weaver
@ 2017-08-08 16:32 ` Andi Kleen
  2017-08-08 16:54 ` Reza Taheri
  2 siblings, 0 replies; 4+ messages in thread
From: Andi Kleen @ 2017-08-08 16:32 UTC (permalink / raw)
  To: H. Taheri; +Cc: linux-perf-users

"H. Taheri" <rezato@gmail.com> writes:

> I am getting 0s, or <not supported>, or <not counted>  for events that are supposed to work on this processor. For example, according to Intel documentation, L2_RQSTS.MISS and L2_RQSTS.REFERENCES are supported on this processor with event code 0x24 and unit masks 0x3f and 0xff. But when I run perf stat (also with -d -d -d to see if the built-in command options can get the cache stats), I get:
>
> [root@prme-perf-broadwell-2 intel]# perf stat -d -d -d -e cycles -e instructions -e r24ff -e r243f sleep 10

You're missing -a, so it's only measuring sleep itself.

-Andi

^ permalink raw reply	[flat|nested] 4+ messages in thread

* Re: Not getting PMC events on Broadwell-EX
  2017-08-08  4:37 Not getting PMC events on Broadwell-EX H. Taheri
  2017-08-08 13:53 ` Vince Weaver
  2017-08-08 16:32 ` Andi Kleen
@ 2017-08-08 16:54 ` Reza Taheri
  2 siblings, 0 replies; 4+ messages in thread
From: Reza Taheri @ 2017-08-08 16:54 UTC (permalink / raw)
  To: linux-perf-users

I see my mistake now. It really is umask followed by eventsel, even if the perf-stat man page says:
       -e, --event=
           Select the PMU event. Selection can be:
           •   a symbolic event name (use perf list to list all events)
           •   a raw PMU event (eventsel+umask) in the form of rNNN where NNN is a hexadecimal event descriptor.

I ran the following to collect the programmable version of retired_inst and unhalted core_cycles, and the outputs match the fixed counters.

[root@prme-perf-broadwell-2 ~]# perf stat -a -e cycles -e instructions -e r03c -e r00c0 sleep 10

 Performance counter stats for 'system wide':

     2,505,518,503      cycles                                                      
       925,145,561      instructions              #    0.37  insn per cycle         
     2,526,621,584      r03c                                                        
       929,277,772      r00c0                                                       

      10.001461051 seconds time elapsed

Thanks,
Reza


^ permalink raw reply	[flat|nested] 4+ messages in thread

end of thread, other threads:[~2017-08-08 16:54 UTC | newest]

Thread overview: 4+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2017-08-08  4:37 Not getting PMC events on Broadwell-EX H. Taheri
2017-08-08 13:53 ` Vince Weaver
2017-08-08 16:32 ` Andi Kleen
2017-08-08 16:54 ` Reza Taheri

This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.