Re: [tip:perf/core] perf/x86/intel/P4: Robistify P4 PMU types

[Borislav Petkov]

On Fri, Apr 26, 2013 at 08:46:52PM +0400, Cyrill Gorcunov wrote:
> On Fri, Apr 26, 2013 at 06:39:52PM +0200, Borislav Petkov wrote:
> >                                              #    0.25  stalled cycles per insn
> >            590,855 branches                  #  237.877 M/sec
> >             12,882 branch-misses             #    2.18% of all branches
> > 	    ^^^^^^^^^^^^^^^^^^^^
> > 
> > Those above?

Well, perf top looks ok to me, here's a snapshot:

   PerfTop:      63 irqs/sec  kernel:79.4%  exact:  0.0% [4000Hz cycles],  (all, 2 CPUs)
--------------------------------------------------------------------------------------------------------------------------------------------

    11.21%  [kernel]               [k] __lock_acquire                
     7.87%  libc-2.13.so           [.] 0x00078b0c                    
     5.78%  libz.so.1.2.7          [.] 0x00003731                    
     4.29%  libpthread-2.13.so     [.] pthread_rwlock_unlock         
     3.74%  libpthread-2.13.so     [.] pthread_rwlock_rdlock         
     3.67%  [kernel]               [k] lock_release                  
     2.55%  [kernel]               [k] lock_acquire                  
     2.27%  perf                   [.] symbols__insert               
     2.15%  sshd                   [.] 0x0004707e                    
     1.62%  libc-2.13.so           [.] vfprintf                      
     1.58%  [kernel]               [k] mark_held_locks               
     1.40%  [kernel]               [k] do_raw_spin_lock              
     1.37%  [kernel]               [k] trace_hardirqs_on_caller      
     1.29%  [kernel]               [k] sub_preempt_count             
     1.17%  perf                   [.] symbol_filter                 
     1.13%  [kernel]               [k] mark_lock                     
     1.05%  [kernel]               [k] trace_hardirqs_off_caller     
     0.96%  perf                   [.] rb_next                       
     0.94%  libc-2.13.so           [.] memchr                        
     0.80%  libbfd-2.22-system.so  [.] 0x000bb009                    
     0.72%  [kernel]               [k] __schedule                    
     0.71%  [kernel]               [k] ioread16                      
     0.67%  [kernel]               [k] _raw_spin_unlock_irqrestore   
     0.66%  [kernel]               [k] __switch_to                   
     0.59%  [kernel]               [k] do_raw_spin_unlock            
     0.56%  perf                   [.] dso__load_sym
...

I can annotate symbols and disassemble works fine too, along with
refresh and per-insn overhead.

The other trivial test passes too, although branch-misses doesn't get
counted:

./perf stat sleep 1

 Performance counter stats for 'sleep 1':

          1.433368 task-clock                #    0.001 CPUs utilized          
                 1 context-switches          #    0.698 K/sec                  
                 0 cpu-migrations            #    0.000 K/sec                  
               147 page-faults               #    0.103 M/sec                  
            78,446 cycles                    #    0.055 GHz                    
                 0 stalled-cycles-frontend   #    0.00% frontend cycles idle   
                 0 stalled-cycles-backend    #    0.00% backend  cycles idle    [27.37%]
         1,268,044 instructions              #   16.16  insns per cycle         [27.37%]
           223,742 branches                  #  156.095 M/sec                   [27.37%]
     <not counted> branch-misses

       1.002191045 seconds time elapsed

However, if I do this, it works:

./perf stat -e branch-misses sleep 1

 Performance counter stats for 'sleep 1':

             8,583 branch-misses

       1.001992384 seconds time elapsed


Oh, btw, tip/master has

commit 697dfd884438058b15032b0169887c742704434a
Merge: 0fbd06761f5c f697036b93aa
Author: H. Peter Anvin <hpa@xxxxxxxxxxxxxxx>
Date:   Thu Apr 25 14:00:22 2013 -0700

    Merge tag 'efi-urgent' into x86/urgent

as its top commit.

HTH.

-- 
Regards/Gruss,
    Boris.

Sent from a fat crate under my desk. Formatting is fine.
--
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/