On Wed, Mar 20, 2013 at 01:58:58PM -0700, Andrew Morton wrote: > On Tue, 19 Mar 2013 13:50:57 -0700 (PDT) Hugh Dickins <hughd@xxxxxxxxxx> wrote: > > > I find it a bit confusing that we now have these two different clustering > > strategies in scan_swap_map(), one for SSD and one for the rest; and it's > > not immediately obvious what's used for what. > > Yes, having two separation allocation paths is bad and we should work > to avoid it, please. Sooner rather than later (which sometimes never > comes). > > We have a few theories about how the SSD code will worsen things for > rotating disks. But have those theories been tested? Any performance > results? If regressions *are* observed, what is the feasibility of > fixing them up? The problem is I don't know which workload is proper to measure the cluster change impact for rotating disks. That would only happen when swap is fragmented but not too much. I'd assume the impact isn't big, but who knows. If you have something I can test, I'm happy to do. Will address other issues soon. Thanks, Shaohua -- To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to majordomo@xxxxxxxxx. For more info on Linux MM, see: http://www.linux-mm.org/ . Don't email: <a href=mailto:"dont@xxxxxxxxx";> email@xxxxxxxxx </a>