Hi, Vince is correct. The limitation is in the perf_events interface itself. That is why I have the event commented out. You need to find a way to encode the event and then have the kernel decode and use the extra MSR. That is not there yet. I had planned to fix this but got caught up with other issues. Will try to get this fixed.
On Fri, Jul 9, 2010 at 4:49 PM, Vince Weaver <vweav...@eecs.utk.edu> wrote: > On Fri, 9 Jul 2010, Ankush Gupta wrote: > >> Hi, >> >> >> I am interested in measuring per-core remote read memory bandwidth on an >> Intel Nehalem machine >> using libpfm4. >> >> I think the only way to do it is using the OFFCORE_RESPONSE_0 event. Looking >> at the source >> code of libpfm4: intel_nhm_events.h, the event is commented out. The >> encoder for the function >> pfm_nhm_offcore_encode in file ./pfmlib_intel_nhm.c has also been commented >> out. Seems it is >> not supported yet. >> >> Is this a limitation of libpfm4 or perf_events? > > >From what I understand, using OFFCORE_RESPONSE requires access to an > additional processor MSR (one not used by other counters) and that > perf_events currently does not support this. > > Vince > ------------------------------------------------------------------------------ > This SF.net email is sponsored by Sprint > What will you do first with EVO, the first 4G phone? > Visit sprint.com/first -- http://p.sf.net/sfu/sprint-com-first > _______________________________________________ > perfmon2-devel mailing list > perfmon2-devel@lists.sourceforge.net > https://lists.sourceforge.net/lists/listinfo/perfmon2-devel > > ------------------------------------------------------------------------------ This SF.net email is sponsored by Sprint What will you do first with EVO, the first 4G phone? Visit sprint.com/first -- http://p.sf.net/sfu/sprint-com-first _______________________________________________ perfmon2-devel mailing list perfmon2-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/perfmon2-devel
