On Sat, Oct 12, 2013 at 9:42 PM, Nicolas Pitre <nico@xxxxxxxxxxx> wrote: > On Fri, 11 Oct 2013, Duy Nguyen wrote: > >> On Fri, Oct 11, 2013 at 07:22:59PM +0700, Duy Nguyen wrote: >> > > > Maybe we could make an exception and allow the tree walker to pass >> > > > pv4_tree_cache* directly to decode_entries so it does not need to do >> > > > the first lookup every time.. >> > > > >> > > > Suggestions? >> >> Looking at decode_entries() traces I think the "one decode_entries() >> for one tree_entry()" just amplifies the delta chain effect. If you >> hide 3 entries behind 5 layers of copy sequences >> (i.e. tree1->tree2->..->tree5->real-tree-entry), then every >> decode_entries(count=1) will have to go through 5 layers. > > Calling decode_entries() for every tree entry is a bad approach. We > should really implement this as a state machine preserving the entire > state between entries so that moving to the next entry is just a matter > of advancing a pointer in most cases. Yeah. Maybe first step is avoid recursion in decode_entries(). We need to do that eventually to avoid stack overlow due to long depths. -- Duy -- To unsubscribe from this list: send the line "unsubscribe git" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html