Dave, On Mon, Jun 12, 2006 at 08:13:14PM -0400, Dave Nellans wrote: > > I think it is worth upgrading each time there is a new release because > > I fix bugs and cleanup the code. As for the user API, I am trying to > > stabilize it at this point. > > what is the status of the in-kernel API for accessing the per process > data without going out to libpfm? i found some mentions of it in the > mailing list archive. i was hoping to utilize it for some thread > co-scheduling experiments in the next few months. >
The in-kernel API (KAPI) does not support per-thread measurement at this point. You can do per-cpu (system-wide). As you can imagine the difficulty for per-thread from inside the kernel is to stop the thread to attach the context, you do not have the luxury of ptrace(). -- -Stephane _______________________________________________ perfmon mailing list [email protected] http://www.hpl.hp.com/hosted/linux/mail-archives/perfmon/
