Re: performance counters on core-duo

Tech-Archive recommends: Repair Windows Errors & Optimize Windows Performance

Yes, the result of 50% may mean 100% of the first processor and 0% of the
second , but OTOH it may be 50% of first and second simultaneously
Arkady

Henry Markov" <hm@xxxxxxxxxxx> wrote in message
news:OOqRlFU1HHA.1168@xxxxxxxxxxxxxxxxxxxxxxx
I'm retrieving the counters from the registry similar to:
http://msdn2.microsoft.com/en-us/library/aa373178.aspx
I know this can be complicated involving scaling the values etc. but when
first developed all results were verified against the Performance tool in
Administrative Tools. The results for the core-duo CPU seem exactly right
if the assumption that the per-process values are relative to only one
core is correct.

HM

"Skywing [MVP]" <skywing_NO_SPAM_@xxxxxxxxxxxxxxxxxxx> wrote in message
news:%23JQa8PT1HHA.3536@xxxxxxxxxxxxxxxxxxxxxxx
What APIs are you using? Are you talking about the processor performance
counters (e.g. rdtsc) or OS APIs?

--
Ken Johnson (Skywing)
Windows SDK MVP
http://www.nynaeve.net
"Henry Markov" <hm@xxxxxxxxxxx> wrote in message
news:eD1vOqR1HHA.1188@xxxxxxxxxxxxxxxxxxxxxxx
I programmatically get CPU and some per-process performance counters for
my application that I have migrated from a single to dual core machine.
For the entire system my CPU utilization dropped about in half on the
core-duo which is in line with expectations. However for two processes
that account for most of the system CPU utilization, each process
utilization is about the same as on the old machine and the sum of the
two exceeds the system utilization.

I make sense out of this by assuming that the system utilization is
reported against the capacity of both cores but per process utilization
is reported only against the host core. Is this the right way to
interpret what I see?

HM






.

Relevant Pages

  • Re: performance counters on core-duo
    ... I'm retrieving the counters from the registry similar to: ... my application that I have migrated from a single to dual core machine. ... For the entire system my CPU utilization dropped about in half on the ... I make sense out of this by assuming that the system utilization is ...
    (microsoft.public.win32.programmer.kernel)
  • Re: performance counters on core-duo
    ... Are you talking about the processor performance counters or OS APIs? ... However for two processes that account for most of the system CPU utilization, each process utilization is about the same as on the old machine and the sum of the two exceeds the system utilization. ... I make sense out of this by assuming that the system utilization is reported against the capacity of both cores but per process utilization is reported only against the host core. ...
    (microsoft.public.win32.programmer.kernel)
  • cpufreq_ondemand governor - problem
    ... on dualcore AMD - if you use cpufreq_ondemand governor and your program is one_process/one_thread intensive one core is busy and second is doing nothing - governor is droping speed on both cores to lowest speed - slowing down busy core process - my dualcore-AMD do this i'm not shure if it is only AMD or INTEL problem too ... core1 - utilization 100% speed lowest possible ...
    (Linux-Kernel)
  • Re: multithreading for pathfinding?
    ... core machine with a radeon x800 pci-express graphics adapter. ... windowed = 10-20% utilization ... fullscreen = 53% util ... window func is passing WM_PAINT through to DefWindowProc. ...
    (microsoft.public.win32.programmer.directx.graphics)
  • cpufreq ondemand governor problem on SMP DUALCORE AMD
    ... kill one - BAD now one core utilization ~100% but cpu speed dropped to core with lowest utilization - fluctuating from min at this level depending on system activity. ... any possibility to setup ondemand governor for max speed if only 1 core utilization is maxed - full speed for single-process/single-thread activity? ...
    (Linux-Kernel)