On Tue, 2013-09-24 at 13:07 -0700, Josh Stone wrote: > On 09/24/2013 01:53 AM, Mark Wielaard wrote: > > Hopefully with your suggestions for some improvements in the DIE > > classifying we can get rid of the last remaining "slowdown". > > We can't expect to get rid of all slowdown, because it is doing more > analysis of the tree regardless.
yeah :) Well, we can always hope. But yes, given that before we didn't even try to find everything by following imported/child DIEs, it will be hard to get no slowdown at all indeed. > But anyway, here's a quick patch to "unclassify" visit_scopes. This is > good for a pretty consistent 30-40ms savings on my machine. Nothing > earth-shattering, but it's some recouped loss. > > If you approve, I can send it more formally. (w/ ChangeLog, etc.) Yes, I approve. It simplifies the code, which obviously was doing too much work. Faster and simpler, while not changing the actual logic underlying the recursion sounds good to me. Thanks, Mark
