[perfmon2] Interpreting output with and without long sampling period

g the numbers incorrectly? Detailed command line and output is below. This is on a Nehalem processor. --Ananth pfmon-3.8/pfmon $ ./pfmon --smpl-module=compact -eINSTRUCTIONS_RETIRED,L2_LINES_IN:ANY,LAST_LEVEL_CACHE_MISSES,UNHALTED_REFERENCE_CYCLES --long-smpl-periods=3,3000,2000 --no-cmd-output

[perfmon2] Counting Nehalem Uncore events

, PMD21, PMD22, PMD23, PMD24, PMD25, PMD26, PMD27, PMD28) I am unable to figure out what is going wrong, since I am specifying only 1 event and only 1 core. There is only 1 CPU socket (socket 0). I am using pfmon version 3.8. Any suggestions? -- Ananth

Re: [perfmon2] Counting Nehalem Uncore events

(10Hz) host kernel architecture: x86_64 $cat /sys/kernel/perfmon/pmu_desc/model Intel architectural -- Ananth On Fri, Apr 17, 2009 at 5:02 PM, stephane eranian wrote: > What does pfmon -I report? > > Make sure you have the right PMU module inserted. > /sys/kernel/perf

Re: [perfmon2] Counting Nehalem Uncore events

Hi Stephane, I have CONFIG_X86_PERFMON_INTEL_NHM=y. So it will get built into the kernel. Despite that, the right PMU module is not being loaded. I have CONFIG_X86_PERFMON_INTEL_ARCH=y too. Is that causing a conflict? -- Ananth On Sat, Apr 18, 2009 at 2:14 AM, stephane eranian wrote: >

Re: [perfmon2] Counting Nehalem Uncore events

Thanks Stephane. Works fine. -- Ananth On Mon, Apr 20, 2009 at 7:18 AM, stephane eranian wrote: > On Sun, Apr 19, 2009 at 2:21 AM, stephane eranian > wrote: >> On Sat, Apr 18, 2009 at 6:04 PM, Ananth wrote: >>> Hi Stephane, >>> I have CONFIG_X86_PERFMON_INT

[perfmon2] Performance MSRs used by perfmon

there a way we can force perfmon to use only the PERF_EVT_SEL MSRs? - Ananth -- Enter the BlackBerry Developer Challenge This is your chance to win up to $100,000 in prizes! For a limited time, vendors submitting new