Interpret cahce events on multicore, shared cache processor

[Joe H. <joe...@gm...>]

Hi
I'm working on a Intel Core i7 (Nehalem) box with 4 cores and
hyper-threading enabled. For this kind of multicore with shared cache
processor, how to properly interpret LLC_MISS events in oprofile
report?

Here are the commands I issued:
opcontrol --start --event=LLC_MISSES:10000 --separate=cpu
opreport

part of the report
START
--------------
CPU: Intel Core/i7, speed 1197 MHz (estimated)
Counted LLC_MISSES events (Last level cache demand requests from this
core that missed the LLC) with a unit mask of 0x41 (No unit mask)
count 10000
Samples on CPU 0
Samples on CPU 1
Samples on CPU 2
Samples on CPU 3
Samples on CPU 4
Samples on CPU 5
Samples on CPU 6
Samples on CPU 7
            cpu:0|            cpu:1|            cpu:2|
cpu:3|            cpu:4|            cpu:5|            cpu:6|
 cpu:7|
  samples|      %|  samples|      %|  samples|      %|  samples|
%|  samples|      %|  samples|      %|  samples|      %|  samples|
 %|
------------------------------------------------------------------------------------------------------------------------------------------------
      130 28.9532        11  2.4444        24 15.3846         2
2.3529         0       0        67 32.8431        10  5.0761         3
11.5385 opera

-------------
report ENDS

In the report LLC_MISSES events are recorded on all 8 logical
processors while on Nehalem 3rd level cache (the LLC on Nehalem) is
shared by all logical processors. So is there duplication in the
number of samples? or I'm not supposed to use --separate=cpu ?

Thanks.

-- 
Joe H