On Wed July 11 2007 5:42 pm, Stephane Eranian wrote: > On Wed, Jul 11, 2007 at 05:28:01PM -0500, Kevin Corry wrote: > > I happened to be reading through the kernel routine > > __pfm_ctxswout_thread() in perfmon/perfmon_ctxsw.c and noticed that the > > if (need_save_pmds) > > pfm_save_pmds(ctx, set); > > is inside the "#ifdef CONFIG_SMP". It seems like that should be outside > > the "#ifdef CONFIG_SMP". Can we really assume that the counters do not > > need to be saved on a UP system? Or is it implicitly assuming that the > > pfm_arch_ctxswout_thread() call will have saved the PMDs in that case? If > > so, I don't think that the PowerPC code is handling that correctly. Any > > thoughts? > > In UP mode, we use a lazy save/restore approach. On ctxswout, we stop > monitoring but we do not save, hoping we'll find our state back next time > around. In case of conflict, the old state is saved by the newcomer. AFAIK, > there should not be any need for arch-specific code for this.
Ok, I see it now. Thanks for the explaination. -- Kevin Corry [EMAIL PROTECTED] http://www.ibm.com/linux/ _______________________________________________ perfmon mailing list [email protected] http://www.hpl.hp.com/hosted/linux/mail-archives/perfmon/
