Re: [PerfSuite-users] [Ptools-perfapi] Question Papi, how to use papi to profile a real type of daemon process?

[Rick K. <rk...@nc...>]

Dawei,

I'm glad to hear you are making a little progress.  Regarding the reason=20
for the zeroes appearing in your later tests: if you examine the output of=
=20
psprocess you will notice that it indicates that multiplexing was enabled=
=20
for the run with zeroes, but not for the run that returned actual event=20
counts.  I believe that this is the underlying reason that zeroes were=20
reported.

On an Itanium 2 system such as you are using, there are only four actual=20
registers available for the performance counters, which limits the number=
=20
of events that can be counted at any given time.  When you supply a=20
configuration file that requires more events than the underlying hardware=
=20
supports, PerfSuite switches into multiplexed operation (through PAPI).=20
What that means is that the available registers are "timeshared" among the=
=20
events: first one event is counted for a short period of time, then the=20
next, then the third, and so on.  At the end of the measurement, the=20
events that were accumulated are scaled up accordingly and that is what is=
=20
reported as the final result.  This is an approximation, and for=20
longer-running programs is usually quite good, but for shorter-running=20
programs it can yield unexpected results.  In the extreme, it is possible=
=20
that the program being measured completed before one or more of the=20
specified events had a chance to be "made active" even once.

I believe the current implementation of multiplexing in PAPI only allows=20
for one event being active in any timeslice, even if the processor=20
actually supports more than one.

So my guess is that what is occurring here is that your program is not=20
running for long enough to accumulate any event counts when multiplexing=20
is enabled.  The run that did not report zeroes only ran for ~36,000=20
cycles, which is much less than a second on your system.

My recommendation is to limit your runs to those that use configurations=20
for which multiplexing is not required.  This should at least give results=
=20
that are greater than zero.  Whether or not runs of these lengths provide=
=20
information that is useful for analysis is another matter, but that's a=20
judgement to be made by the end user.

There is currently no command-line utility in PerfSuite that allows one to=
=20
query a given configuration file to learn if it would require multiplexing=
=20
(this would be a useful addition, I think), so the easiest way is to=20
experiment with commands like "ls" in combination with test configuration=
=20
files and examine the output to see if multiplexing was used.  Of course,=
=20
it's good to know how many registers are available on your system in=20
advance, as that lets you know the maximum, but you still have to do test=
=20
runs because some PAPI events ("derived events") are actually composed of=
=20
more than one underlying native event.

Rick

On Fri, 12 Jan 2007, liudawei wrote:

[ ... ]

>  What is the problem happen ? Waiting your reply.
>
>
>  Best Reagard
>  Dawei Liu
>  Renmin University of China
> 100872 Beijing,China
>
>
> ---------------------------------
> =C7=C0=D7=A2=D1=C5=BB=A2=C3=E2=B7=D1=D3=CA=CF=E4-3.5G=C8=DD=C1=BF=A3=AC20=
M=B8=BD=BC=FE=A3=A1