Date: Wed, 28 Sep 2005 13:07:11 -0400 From: John Baldwin <jhb@FreeBSD.org> To: Jeremie Le Hen <jeremie@le-hen.org> Cc: cvs-src@FreeBSD.org, src-committers@FreeBSD.org, cvs-all@FreeBSD.org Subject: Re: cvs commit: src/sys/kern kern_prot.c vfs_export.c src/sys/nfsserver nfs_srvsock.c nfs_srvsubs.c src/sys/sys ucred.h src/sys/ufs/ufs ufs_vnops.c Message-ID: <200509281307.13026.jhb@FreeBSD.org> In-Reply-To: <20050928091638.GB2151@obiwan.tataz.chchile.org> References: <200509271809.j8RI9gCO020672@repoman.freebsd.org> <20050928091638.GB2151@obiwan.tataz.chchile.org>
next in thread | previous in thread | raw e-mail | index | archive | help
On Wednesday 28 September 2005 05:16 am, Jeremie Le Hen wrote: > Hi John, > > > jhb 2005-09-27 18:09:42 UTC > > > > FreeBSD src repository > > > > Modified files: > > sys/kern kern_prot.c vfs_export.c > > sys/nfsserver nfs_srvsock.c nfs_srvsubs.c > > sys/sys ucred.h > > sys/ufs/ufs ufs_vnops.c > > Log: > > Use the refcount API to manage the reference count for user credentials > > rather than using pool mutexes. > > Regarding your great job on atomic operations and the modification you > made on process argument and process limit structures as well as > user credentials, I wonder if you managed to get some figures > illustrating the improvement. No, I didn't run any before/after benchmarks, just stability tests to make sure I didn't break anything. I'm pretty sure 2 atomic operations aren't cheaper than just 1 though. However, on some archs (ppc and ia64 currently) there might be a penalty in the edge case where the atomic_cmpset loop races and has to do more than one atomic_cmpset. However, in that case it would have had to contest on the mutex anyways, so even then I still think it would be a win. -- John Baldwin <jhb@FreeBSD.org> <>< http://www.FreeBSD.org/~jhb/ "Power Users Use the Power to Serve" = http://www.FreeBSD.org
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?200509281307.13026.jhb>