Need Help!!! Total CPU Time Utilization Mismatch in Vtune Report

[Sidd_p_]

Hi,

   Please find the V-Tune Report as a attachment.

i am confused with this Report. Let me confirm my understanding from you guys. Please reply.

V-Tune showing me that cmdLoader_c() this functions is taking 3.726 s CPU time.

Callees                                 CPU Time: Total by Utilization    CPU Time: Self by Utilization
cmdLoader_c::cmdLoader_c    3.726s        3.034s                     100.0%    0.016s

Now if i expand this i get following CPU times.

[mscorlib.ni.dll]                          0.526s    0.028s                         20.1%    0s
uiDveSheet_c::initDve                 0.091s    0.001s                         6.4%    0s
[PresentationCore.ni.dll]              0.037s    0s                               3.3%    0s
[PresentationFramework.ni.dll]     0.021s    0.003s                         6.6%    0s
[WindowsBase.ni.dll]                  0.006s    0.001s                          0.0%    0s
uiBaseCommand_c::run              0.003s    0.002s                         0.0%    0s
uiDvePage_c::uiDvePage_c         0.003s    0.003s                         0.0%    0s
uiBlockMan_c::pushBlockState    0.002s    0.001s                        2.6%    0.001s
uiContourManager_c::endContourSelection    0.002s                     0s    0.0%    0s
`anonymous namespace'::newDeleteFace    0.001s                       0.001s    0.0%    0s

now my question is if do Total of these expanded CPU times it will not be 3.726 s. i want to know which function is taking 3 secs CPU Time. please find attachment if you didn't get my question..please reply.

[Peter_W_Intel]

Thank you of reporting this!

I just want to know if you set Call Stack Mode: "User/System function". If there was no system functions missed, can you provide result directory, or report this to https://premier.intel.com.

Do you use VTune(TM) Amplifier XE 2015?

[Sidd_p_]

Thanks Peter for update.

I am Using VTune(TM) Amplifier XE 2013. can you tell me where should i verify that Call Stack Mode. I am running Basic Hotspots Algorithm Analysis and there i am not seeing any Call Stack Mode. 

-Sidd.

[Bernard]

Are those functions below the cmdLoader_c::cmdLoader_c  calles of cmdLoader_c ?

[Peter_W_Intel]

Sidd p. wrote:

Thanks Peter for update.

I am Using VTune(TM) Amplifier XE 2013. can you tell me where should i verify that Call Stack Mode. I am running Basic Hotspots Algorithm Analysis and there i am not seeing any Call Stack Mode. 

-Sidd.

It should be in your bottom-up report, see the corner of left-bottom. 

[Dmitry_P_Intel1]

Hello Sidd,

Per what I see from the picture and you report published we have following:

cmdLoader_c::cmdLoader_c has 3.726s of Total CPU time and 3.034s of Self CPU time so all callees that ware called from this function consumed 0.692 in sum. 3.034s was consumed by cmdLoader_c::cmdLoader_c itself - you can double click on the grid row of the function to see the function source and assembly and disctribution of 3.034s by the function source line or assembly instructions.

Thanks & Regards, Dmitry

[Bernard]

c>>>mdLoader_c::cmdLoader_c has 3.726s of Total CPU time and 3.034s of Self CPU time so all callees that ware called from this function consumed 0.692 in sum. 3.034s was consumed by cmdLoader_c::cmdLoader_c itself >>>

I also came later to the same conclusion.

[Peter_W_Intel]

The problem was that contributions (summary of CPU time) from callers was NOT equal to CPU time of hot function (cmdLoader_c::cmdLoader_c   ), whatever CPU time was Total Time or Self Time. I doubt that some callers are from system function, that was why I asked to switch Call Stack mode. 

Otherwise, this is a bug!