John,
totally acceptable what you say -
I can see the activities of the monitor-program in the first 6
entriesbut then it is the CPU-hog only.
The LPP-functionality is there to distinct samples made during activities.
...so it is IMHO vital to have it accurate.
--
Martin
Pi_cap_CPU - all y
It is common---quasi-obligatory in certain Monte Carlo
computations---to discard an initial set of sampling observations or
treat them as insignificant.
The notion is that what is being observed must reach a steady state
before meaningful measurement is possible and that [for the usual
Heisenberg-
It is common, indeed q
On 10/12/12, Martin Truebner wrote:
> Driving the CPU at 100% I get 26926 samples. This is 10 percent less
> than the 3 I expected, but I should allow a little slack for the
> sampling itself.
>
> But what puzzles me most is the fact that the LPP(*1) I give ahead of
> a
Driving the CPU at 100% I get 26926 samples. This is 10 percent less
than the 3 I expected, but I should allow a little slack for the
sampling itself.
But what puzzles me most is the fact that the LPP(*1) I give ahead of
activating it via LSCTL(*2) is ignored for the first 87 entries. I give
i
Did I name the book:
The Load-Program-Parameter and the CPU-Measurement Facilities
SA23-2260-01
--
Martin
Pi_cap_CPU - all you ever need around MWLC/SCRT/CMT in z/VSE
more at http://www.picapcpu.de
David (et al)
I have found a sentence that might be a pointer to the reason for
the low number of samples.
On page 2.16 first paragraph it says
5. One may observe the wait-state (W) bit in sample
data being set to one for dedicated CPUs, or
for nondedicated CPUs in some unusual situations.
For n
David,
>> Just a guess: are samples collected only when the LPAR is busy?
It does not say so (*1) and there is a bit in each sample that says LPAR
was in wait.
>> And if so, was the LPAR 100% busy?
Zero (nothing- at all) on the machine
>> If not, then I would expect the the number of samples w
Martin,
Just a guess: are samples collected only when the LPAR is busy? And if so,
was the LPAR 100% busy? If not, then I would expect the the number of
samples would be proportional to the usage.
David
Am Montag, 8 Oct 2012 14:54:01 +0200, Martin Truebner schrieb:
>I am a little puzzled.
>
>I a
I am a little puzzled.
I ask the CPU for its speed (QSI) and it comes back with a number (407
cycles per micro second) - I multiply it with 2000 and use the result
in an LSCTL to specify the interval (together with the activation bits
and the structures) (all checked and it is there).
Now I wait
