On Thu, 2007-08-09 at 02:00 -0700, Stephane Eranian wrote: > Hi, > > On Thu, Aug 09, 2007 at 04:52:16PM +0800, Zhang, Yanmin wrote: > > If I use option --long-smpl-period, pfmon reported another error. > > > > [EMAIL PROTECTED] tmp]# pfmon -u -k --no-cmd-output --smpl-module=dear-hist > > --smpl-data-view --smpl-show-function > > --smpl-sort-bycount --smpl-save-raw --resolve-addresses > > --long-smpl-period=10000 -e L1I_EAR_CACHE_LAT8 ./tt > > sampling module only works with one DATA_EAR_* event at a time > > > > The dear-hist module only works for Data EAR events, you are passing an > instruction EAR event. Pfmon does not yet have a sampling module to > present I-EAR data. This should not be too hard to write based on dear-hist. Thanks for your kind help. Now I could use pfmon to collect dear data. It's an useful feature for NUMA optimization.
One more question: pfmon could show dear data from different view. smpl-level-view could show latency and smpl-inst-view could show the instruction address which accesses the data. I see that pfmon could save result. Could pfmon save the raw data and later on I could use pfmon to show the same data from different views? Thanks. -yanmin _______________________________________________ perfmon mailing list [email protected] http://www.hpl.hp.com/hosted/linux/mail-archives/perfmon/
