Vince, On Thu, Nov 18, 2010 at 4:29 PM, Vince Weaver <vweav...@eecs.utk.edu> wrote: > Hello > > I was wondering if anyone on this list has experience with the > HW_INT_RCV event on the Atom (r5100c8:u if using perf). > I had noticed this event a long time ago. It's been around for a while yet I don't think it's used very much.
All my event table sources indicate that 0x00c8 is the correct encoding. Are you getting some counts on Intel Core? On Nehalem/Westmere, it's there too with 0x11d. When I try on NHM, I get sensible results: $ perf stat -e hw_int:rcv -a -C 1 -- taskset -c 1 ./noploop 1 noploop for 1 seconds Performance counter stats for 'taskset -c 1 ./noploop 1': 1279 nhm::HW_INT:RCV:k=1:u=1:e=0:i=0:c=0:t=0 1.000743130 seconds time elapsed > I've been running some tests on both an Atom n270 and an Atom 230 > and always get 0 for the count, even on long-running programs that should > have thousands. > > I'm beginning to wonder if this particular event was just a cut-and-paste > error from the Core2 event list (as it has the exact same description and > event num/umask on Core2 as Atom: C8/00 ). > When I try libpfm4/perf_examples/syst_count -e hw_int_rcv:u:k -d 1 on my Atom 230, I also get zero. So either the encoding is wrong or the event is busted. Let me check on this. ------------------------------------------------------------------------------ Beautiful is writing same markup. Internet Explorer 9 supports standards for HTML5, CSS3, SVG 1.1, ECMAScript5, and DOM L2 & L3. Spend less time writing and rewriting code and more time creating great experiences on the web. Be a part of the beta today http://p.sf.net/sfu/msIE9-sfdev2dev _______________________________________________ perfmon2-devel mailing list perfmon2-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/perfmon2-devel
