Re: [Gems-users] The distribution of "total_misses" in ruby stats output

[Dan Gibson <degibson@xxxxxxxx>]

Sequencer identifiers are:
(m_chip_ptr->getID()*RubyConfig::numberOfProcsPerChip()+m_version)

Regards,
Dan

Lide Duan wrote:

I cout the processor ID in the makeReqeust() functions of SimicsDriver and SimicsProcessor. Some output here:

Processor ID of current request in SimicsDriver::makeRequest() : 15
Processor ID of current request in SimicsProcessor::makeRequest() : 15
Processor ID of current request in SimicsDriver::makeRequest() : 14
Processor ID of current request in SimicsProcessor::makeRequest() : 14
Processor ID of current request in SimicsDriver::makeRequest() : 14
Processor ID of current request in SimicsProcessor::makeRequest() : 14
Processor ID of current request in SimicsDriver::makeRequest() : 14
Processor ID of current request in SimicsProcessor::makeRequest() : 14
Processor ID of current request in SimicsDriver::makeRequest() : 14
Processor ID of current request in SimicsProcessor::makeRequest() : 14
Processor ID of current request in SimicsDriver::makeRequest() : 14
Processor ID of current request in SimicsProcessor::makeRequest() : 14
Processor ID of current request in SimicsDriver::makeRequest() : 6
Processor ID of current request in SimicsProcessor::makeRequest() : 6
Processor ID of current request in SimicsDriver::makeRequest() : 6
Processor ID of current request in SimicsProcessor::makeRequest() : 6
Processor ID of current request in SimicsDriver::makeRequest() : 6
Processor ID of current request in SimicsProcessor::makeRequest() : 6
Processor ID of current request in SimicsDriver::makeRequest() : 13
Processor ID of current request in SimicsProcessor::makeRequest() : 13
Processor ID of current request in SimicsDriver::makeRequest() : 6
Processor ID of current request in SimicsProcessor::makeRequest() : 6

SimicsProcessor also sees the various requests. But I don't know how to reach the processor ID of Sequencer. The input parameter of Sequencer::makeRequest() is "CacheMsg& request" in whose member variables I cannot find the processor ID.

Lide

On 8/24/07, Dan Gibson <degibson@xxxxxxxx> wrote:

Hmmm... try the same for SimicsProcessor and Sequencer.

Regards,
Dan

Lide Duan wrote:
> Hi Dan,
>
> I did what you said, and got the processor IDs of each request. A
> piece of the output is shown here:
>
> Process ID of current request: 14
> Process ID of current request: 9
> Process ID of current request: 9
> Process ID of current request: 9
> Process ID of current request: 9
> Process ID of current request: 9
> Process ID of current request: 5
> Process ID of current request: 9
> Process ID of current request: 5
> Process ID of current request: 9
> Process ID of current request: 5
> Process ID of current request: 9
> Process ID of current request: 9
> Process ID of current request: 13
> Process ID of current request: 5
> Process ID of current request: 13
> Process ID of current request: 5
> Process ID of current request: 13
> Process ID of current request: 5
> Process ID of current request: 13
> Process ID of current request: 13
> Process ID of current request: 5
> Process ID of current request: 13
> Process ID of current request: 5
> Process ID of current request: 13
> Process ID of current request: 5
> Process ID of current request: 13
> Process ID of current request: 13
> Process ID of current request: 5
> Process ID of current request: 5
> Process ID of current request: 5
> Process ID of current request: 6
> Process ID of current request: 5
> Process ID of current request: 5
> Process ID of current request: 6
> Process ID of current request: 6
> Process ID of current request: 6
> Process ID of current request: 6
> Process ID of current request: 2
> Process ID of current request: 2
> Process ID of current request: 2
> Process ID of current request: 10
>
> Seems that Ruby is really seeing the requests from all the processors...
>
> Lide
>
> On 8/24/07, *Dan Gibson* < degibson@xxxxxxxx
> <mailto: degibson@xxxxxxxx>> wrote:
>
>     Interesting. Can you modify SimicsDriver.C to print processor IDs of
>     each request, and verify manually whether Ruby is actually seeing
>     requests from all 16 procs? The processor ID can be found right after
>     the call to SIMICS_get_proc_no() in makeRequest().
>
>     Regards,
>     Dan
>
>     Lide Duan wrote:
>     > I am using Simics2.2.19 + GEMS1.4, and simulating sarek target
>     machine
>     > running Solaris9 OS. I also verified in the Simics command line,
>     there
>     > are only phys_mem0.xxx commands. So I think there is only one
>     > phys_mem, i.e. phys_mem0. Then what might be the problem?
>     >
>     > Lide
>     >
>     > On 8/24/07, *Dan Gibson* <degibson@xxxxxxxx
>     <mailto:degibson@xxxxxxxx>
>     > <mailto: degibson@xxxxxxxx <mailto: degibson@xxxxxxxx>>> wrote:
>     >
>     >     Which version of Simics are you using, and what is your target
>     >     machine?
>     >
>     >     My initial thought is that there are more than one phys_mem*
>     objects,
>     >     and Ruby is only attaching itself to phys_mem0 (in Simics
>     2.x/Sarek
>     >     target, there is _ONLY_ phys_mem0). However, with some other
>     >     versions of
>     >     Simics and/or different target machines, there are sometimes
>     phys_mem
>     >     objects for EACH cpu -- verify that you only have one
>     phys_mem* object
>     >     (via the Simics command line), and that its name is phys_mem0.
>     >
>     >     Regards,
>     >     Dan
>     >
>     >     Lide Duan wrote:
>     >     > I found something strange when looking at the ruby stats
>     output
>     >     files.
>     >     > I am simulating some 16p checkpoints. If I place all the
>     16p on a
>     >     > single chip (g_PROCS_PER_CHIP 16), the results related to
>     cache
>     >     misses
>     >     > are shown as follow:
>     >     >
>     >     > Total_misses: 2537939
>     >     > total_misses: 2537939 [ 2537939 0 0 0 0 0 0 0 0 0 0 0 0 0
>     0 0 ]
>     >     > user_misses: 1763423 [ 1763423 0 0 0 0 0 0 0 0 0 0 0 0 0 0
>     0 ]
>     >     > supervisor_misses: 774516 [ 774516 0 0 0 0 0 0 0 0 0 0 0 0
>     0 0 0 ]
>     >     >
>     >     > I suppose the 16 numbers in the brackets correspond to the
>     >     misses on
>     >     > each processor, but as we can see only misses on the 1st
>     processor
>     >     > were observed. On the other hand, I also got the followings:
>     >     >
>     >     > instruction_executed: 1322569080 [ 82094180 83437607 81021513
>     >     81093394
>     >     > 82632809 122205588 82083483 80671650 81138678 80191395
>     79408095
>     >     > 79763896 81353502 81528474 62244475 81700341 ]
>     >     > cycles_per_instruction: 4.63392 [ 4.66589 4.59077 4.72767
>     4.72348
>     >     > 4.63548 3.13441 4.6665 4.74817 4.72084 4.77661 4.82372
>     4.80221
>     >     4.70837
>     >     > 4.69827 6.15384 4.68839 ]
>     >     > misses_per_thousand_instructions: 1.91895 [ 30.915 0 0 0 0
>     0 0 0
>     >     0 0 0
>     >     > 0 0 0 0 0 ]
>     >     >
>     >     > So definitely all the processors were running something,
>     but why
>     >     were
>     >     > the misses on the 1st processor observed only?
>     >     >
>     >     > To address this problem, I tried different configurations.
>     If I
>     >     place
>     >     > the processors on 4 chips each containing 4p, the first 4
>     numbers in
>     >     > the total_misses brackets are not zeros. Also, if one
>     processor
>     >     on one
>     >     > chip (totally 16 chips), all the 16 numbers are not zeros.
>     >     Therefore,
>     >     > I guess the numbers indicate the misses on each CHIP, not
>     each
>     >     > processor. Am I right? or did I miss something here?
>     Actually I
>     >     tried
>     >     > different workloads with different network topologies, but
>     got the
>     >     > similar results. Can anybody give me some explanation?
>     >     >
>     >     > Thanks,
>     >     > Lide
>     >     >
>     >
>     ------------------------------------------------------------------------
>     >     >
>     >     > _______________________________________________
>     >     > Gems-users mailing list
>     >     > Gems-users@xxxxxxxxxxx <mailto:Gems-users@xxxxxxxxxxx>
>     <mailto: 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.
>     >     >
>     >     >
>     >
>     >     --
>     >     http://www.cs.wisc.edu/~gibson
>     <http://www.cs.wisc.edu/%7Egibson> <http://www.cs.wisc.edu/%7Egibson>
>     >     [esc]:wq!
>     >
>     >     _______________________________________________
>     >     Gems-users mailing list
>     >     Gems-users@xxxxxxxxxxx <mailto: Gems-users@xxxxxxxxxxx>
>     <mailto: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/
>     <https://lists.cs.wisc.edu/archive/gems-users/ >
>     >     <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/
>     <https://lists.cs.wisc.edu/archive/gems-users/ >" to your search.
>     >
>     >
>
>     --
>     http://www.cs.wisc.edu/~gibson < http://www.cs.wisc.edu/%7Egibson>
>     [esc]:wq!
>
>     _______________________________________________
>     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.
>
>

--
http://www.cs.wisc.edu/~gibson [esc]:wq!

_______________________________________________
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.