As I mentioned earlier email, I ran other applications, but it also does
not have kernel stats.
Note that the kernel data is collected, but it does not collect the stat
for kernel instruction.
I found that the memory request is made on ThreadContext::fetch() in
ooo-pipe.cpp.
The request is initialized like below.
--------------------------------------------------------------------------------------------------------------------------------------------
Memory::MemoryRequest *request =
core.memoryHierarchy->get_free_request(core.get_coreid());
assert(request != NULL);
request->init(core.get_coreid(), threadid, physaddr, 0, sim_cycle,
true, 0, 0, Memory::MEMORY_OP_READ);
request->set_coreSignal(&core.icache_signal);
--------------------------------------------------------------------------------------------------------------------------------------------
It makes the ownerRIP_ variable is always 0. The 'is_kernel' function is
based on the ownerRIP value.
Is that right?
Thanks,
Jeongseob Ahn
2014-03-28 1:15 GMT+09:00 avadh patel <[email protected]>:
>
> On Thu, Mar 27, 2014 at 4:41 AM, J Ahn <[email protected]> wrote:
>
>> I am sorry the above link is wrong.
>> This is the correct link.
>>
>>
>> https://www.mail-archive.com/marss86-devel%40cs.binghamton.edu/msg00237.html
>>
>>
>>
>> 2014-03-27 14:11 GMT+09:00 J Ahn <[email protected]>:
>>
>> Hi all,
>>>
>>> When simulating the sample 'fibonacci' application, which is provided by
>>> parsecROI image, I found that there is no stat in L1 instruction cache for
>>> kernel.
>>> I am sure this is not application problem because the same problem
>>> still occurs when I change the application.
>>> (FYI, the user mode stat for L1 icache is well collected.)
>>>
>>> At 2012, the similar question was posted as below link. However, It does
>>> not seems to fix the problem.
>>>
>>> https://www.mail-archive.com/marss86-devel%40cs.binghamton.edu/msg01364.html
>>>
>>> In fibonacci there is near to 0% kernel execution, so how are you
> running the application depends a lot on kernel level runs. One thing you
> can try is to run 'dd' command to write to some file. That should generate
> good amount of kernel activity.
>
> - Avadh
>
>
>>> Thanks,
>>>
>>> Jeongseob Ahn
>>>
>>
>>
>> _______________________________________________
>> http://www.marss86.org
>> Marss86-Devel mailing list
>> [email protected]
>> https://www.cs.binghamton.edu/mailman/listinfo/marss86-devel
>>
>>
>
_______________________________________________
http://www.marss86.org
Marss86-Devel mailing list
[email protected]
https://www.cs.binghamton.edu/mailman/listinfo/marss86-devel
