Re: [Gems-users] Very high cache miss overhead

[Ikhwan Lee <ikhwan@xxxxxxxxxxxxxxx>]

I've had similar issues with abisko/serengeti checkpoints. My guesses are:

1. Some OS background processes are running like what Byn suggested.
2. Your cache is too small to even hold the small OS code/working set.
3. Spurious interrupts are being generated too often.

In my case, simics cpu was running at 75mhz, and interrupts (probably
timer) were being generated too often that cpu was executing OS
(supervisor) code 90% of the time. I could get around the high
supervisor miss rate by increasing the cpu freq to a higher number
(e.g. 2ghz), but this might affect I/O timings as discussed before:
https://lists.cs.wisc.edu/archive/gems-users/2010-February/msg00049.shtml

Ikhwan

On Fri, Apr 16, 2010 at 3:52 AM, Byn Choi <bynchoi1@xxxxxxxxxxxx> wrote:
> That means those are generated by some OS background process. You can either
> let it run functionally for a while until the system becomes idle (then make
> a checkpoint), or reboot the simulated system. I personally never had such
> issue - I always just boot the machine to get an idle system.
>
> Byn
>
> On Apr 15, 2010, at 12:52 PM, Mark Samuelson wrote:
>
>> They are almost all supervisor misses.  What does that mean?
>>
>> Carole-Jean Wu wrote:
>>>
>>> Are the cache misses you saw supervisor or user misses? Carole
>>> On Thu, Apr 15, 2010 at 1:21 PM, Byn Choi <bynchoi1@xxxxxxxxxxxx
>>> <mailto:bynchoi1@xxxxxxxxxxxx>> wrote:
>>>   What if you didn't run anything, i.e. maybe there is some background
>>>   process?
>>>   Byn
>>>   On Apr 15, 2010, at 11:28 AM, Mark Samuelson wrote:
>>>       We are using the MESI_CMP_filter_directory and are getting over
>>>       900000 cache misses in both L1 and L2 even when are test program
>>>       does nothing but sleep.  Is this level of overhead normal?
>>>       Mark
>>>       _______________________________________________
>>>       Gems-users mailing list
>>>       Gems-users@xxxxxxxxxxx <mailto:Gems-users@xxxxxxxxxxx>
>>>       https://lists.cs.wisc.edu/mailman/listinfo/gems-users
>>>       Use Google to search the GEMS Users mailing list by adding
>>>       "site:https://lists.cs.wisc.edu/archive/gems-users/"; to your
>>> search.
>>>   _______________________________________________
>>>   Gems-users mailing list
>>>   Gems-users@xxxxxxxxxxx <mailto:Gems-users@xxxxxxxxxxx>
>>>   https://lists.cs.wisc.edu/mailman/listinfo/gems-users
>>>   Use Google to search the GEMS Users mailing list by adding
>>>   "site:https://lists.cs.wisc.edu/archive/gems-users/"; to your search.
>>> ------------------------------------------------------------------------
>>> _______________________________________________
>>> Gems-users mailing list
>>> Gems-users@xxxxxxxxxxx
>>> https://lists.cs.wisc.edu/mailman/listinfo/gems-users
>>> Use Google to search the GEMS Users mailing list by adding
>>> "site:https://lists.cs.wisc.edu/archive/gems-users/"; to your search.
>>
>> _______________________________________________
>> Gems-users mailing list
>> Gems-users@xxxxxxxxxxx
>> https://lists.cs.wisc.edu/mailman/listinfo/gems-users
>> Use Google to search the GEMS Users mailing list by adding
>> "site:https://lists.cs.wisc.edu/archive/gems-users/"; to your search.
>>
>
> _______________________________________________
> Gems-users mailing list
> Gems-users@xxxxxxxxxxx
> https://lists.cs.wisc.edu/mailman/listinfo/gems-users
> Use Google to search the GEMS Users mailing list by adding
> "site:https://lists.cs.wisc.edu/archive/gems-users/"; to your search.
>
>