Thanks for the review! > On Jul 1, 2016, at 9:08 PM, Ben Pfaff <b...@ovn.org> wrote: > > On Tue, Jun 21, 2016 at 05:18:45PM -0700, Jarno Rajahalme wrote: >> PMD threads use pvectors but do not need the overhead of the >> concurrent version. Expose the internal non-concurrent version for >> that use. >> >> Signed-off-by: Jarno Rajahalme <ja...@ovn.org> > > Seems like a good idea, if we have a potential user (do we?). >
I just posted a v2 that includes the user in dpif-netdev.c. > I guess that the following comments are actually relevant with or > without this patch. > I added separate patches for these > Any reason why PVECTOR_EXTRA_ALLOC is in pvector.h instead of pvector.c? > > Any reason we couldn't do the following, instead of null-ing in-place? > I think that we have to sort the vector either way, and then we could > get rid of the INT_MIN special case. > I had to slightly change the semantics of PVECTOR_FOR_EACH_PRIORITY to do this. > diff --git a/lib/pvector.c b/lib/pvector.c > index 00880bd..55e5351 100644 > --- a/lib/pvector.c > +++ b/lib/pvector.c > @@ -158,10 +158,7 @@ pvector_impl_remove(struct pvector_impl *impl, void *ptr) > > index = pvector_impl_find(impl, ptr); > ovs_assert(index >= 0); > - /* Now at the index of the entry to be deleted. > - * Clear in place, sort will clean these off. */ > - impl->vector[index].ptr = NULL; > - impl->vector[index].priority = INT_MIN; > + impl->vector[index] = impl->vector[--impl->size]; > } > > void > > Acked-by: Ben Pfaff <b...@ovn.org> I also renamed pvector to cpvector and pvector_impl to pvector. It is pretty straightforward, but I'd like you to scan the changes once more, so I posted a v2. Jarno _______________________________________________ dev mailing list dev@openvswitch.org http://openvswitch.org/mailman/listinfo/dev
