Messages in this thread | | | Date | Wed, 22 Jul 2009 21:17:56 +0800 | Subject | Re: [tip:core/locking 1/3] lockdep:reintroduce generation count to make BFS faster | From | Ming Lei <> |
| |
2009/7/22 Peter Zijlstra <a.p.zijlstra@chello.nl>: > On Wed, 2009-07-22 at 20:58 +0800, tom.leiming@gmail.com wrote: >> From: Ming Lei <tom.leiming@gmail.com> >> >> We still can apply DaveM's generation count optimization to >> BFS, so reintroduce it. > > Please elaborate, changelogs are important.
Sorry, I'll resend the patches and provide a elaborate changlogs.
> > You fail to explain why it still applies and what the gains are. > > (I can guess, but you explaining makes sure we're on the same page) >
gains: 1), avoid to allocate the bfs_accessed array;
2), remove the bitmap_zero(bfs_accessed, MAX_LOCKDEP_ENTRIES); in each BFS, which is a little time consuming since MAX_LOCKDEP_ENTRIES is very large.(16384UL)
-- Lei Ming
| |