Yes, it worked. I guess I had wrongly interpreted how task works. Shouldn't
it only count the events from the passed binary and it's child processes
(if inheritance flag is turned on)?
On Tue, Sep 23, 2014 at 9:26 PM, Stephane Eranian <eran...@googlemail.com>
wrote:
> On Tue, Sep 23, 2014 at 8:58 PM, Bhavishya Goel
> <bhavishya.g...@gmail.com> wrote:
> > Hi Vince,
> >
> > I am using lipfm4 directly. This is the command line I use:
> >
> > task -i -e MEM_UOPS_RETIRED:ALL_LOADS,MEM_UOPS_RETIRED:ALL_STORES ./loop.
>
> task -i -e MEM_UOPS_RETIRED:ALL_LOADS:u,MEM_UOPS_RETIRED:ALL_STORES:u
> ./loop.
>
> See if this helps.
>
>
> > From what I understand, the task example in libpfm4 doesn't count kernel
> > level events.
> >
> > Regards
> > Bhavi
> >
> > On Tue, Sep 23, 2014 at 8:24 PM, Vince Weaver <vincent.wea...@maine.edu>
> > wrote:
> >>
> >>
> >>
> >>
> >> On Tue, 23 Sep 2014, Bhavishya Goel wrote:
> >>
> >> > When I use Vtune to verify these values, it counts 0 loads and stores
> >> > for
> >> > this loop as expected.
> >>
> >> you need to specify you only want userspace measurements.
> >> By default you are probably getting kernel values included too.
> >> You don't give your method for measuring things, but if you are using
> >> perf for example you would want
> >> instructions:u
> >> which is user instructions only, not just
> >> instructions
> >> which gives you kernel and user.
> >>
> >> If you're using libpfm4 directly you can set this at event creation
> time.
> >>
> >> Vince
> >>
> >>
> >>
> >> >
> >> > On Tue, Sep 23, 2014 at 7:06 PM, Bhavishya Goel
> >> > <bhavishya.g...@gmail.com>
> >> > wrote:
> >> > Hi,
> >> > I am using the task example to count loads and stores for my
> >> > application, which mainly constitutes of a loop doing some integer
> >> > additions. The assembly of the loop looks like this:
> >> >
> >> > .L3:
> >> > addl %eax, %ebx
> >> > addl %ecx, %edx
> >> > addl %esi, %edi
> >> > addl $1, %ebp
> >> > cmpl $1999999999, %ebp
> >> > jle .L3
> >> >
> >> >
> >> > This loop generates around 300K loads (MEM_UOPS_RETIRED:ALL_LOADS) and
> >> > 200K stores (MEM_UOPS_RETIRED:ALL_STORES) as per task even though
> >> > there are no memory operations in this loop. What can possibly
> >> > generate these loads and stores? Could it have something to do with
> >> > context switching?
> >> >
> >> > I am using an Ivy Bridge machine (HT turned off) with linux kernel
> >> > 3.16.2 and libpfm 4.5.0.
> >> >
> >> > Regards
> >> > Bhavi
> >> >
> >> > --
> >> > ಠ_ಠ
> >> >
> >> >
> >> >
> >> >
> >> > --
> >> > ಠ_ಠ
> >> >
> >> >
> >>
> >> Vince Weaver
> >> vincent.wea...@maine.edu
> >> http://www.eece.maine.edu/~vweaver/
> >
> >
> >
> >
> > --
> > ಠ_ಠ
> >
> >
> ------------------------------------------------------------------------------
> > Meet PCI DSS 3.0 Compliance Requirements with EventLog Analyzer
> > Achieve PCI DSS 3.0 Compliant Status with Out-of-the-box PCI DSS Reports
> > Are you Audit-Ready for PCI DSS 3.0 Compliance? Download White paper
> > Comply to PCI DSS 3.0 Requirement 10 and 11.5 with EventLog Analyzer
> >
> http://pubads.g.doubleclick.net/gampad/clk?id=154622311&iu=/4140/ostg.clktrk
> > _______________________________________________
> > perfmon2-devel mailing list
> > perfmon2-devel@lists.sourceforge.net
> > https://lists.sourceforge.net/lists/listinfo/perfmon2-devel
> >
>
--
ಠ_ಠ
------------------------------------------------------------------------------
Meet PCI DSS 3.0 Compliance Requirements with EventLog Analyzer
Achieve PCI DSS 3.0 Compliant Status with Out-of-the-box PCI DSS Reports
Are you Audit-Ready for PCI DSS 3.0 Compliance? Download White paper
Comply to PCI DSS 3.0 Requirement 10 and 11.5 with EventLog Analyzer
http://pubads.g.doubleclick.net/gampad/clk?id=154622311&iu=/4140/ostg.clktrk
_______________________________________________
perfmon2-devel mailing list
perfmon2-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/perfmon2-devel
