* Re: + fs-bufferc-make-bh_lru_install-more-efficient.patch added to -mm tree [not found] ` <alpine.DEB.2.20.1706011050590.8835@east.gentwo.org> @ 2017-06-03 5:44 ` Eric Biggers 2017-06-05 13:51 ` Christoph Lameter 0 siblings, 1 reply; 2+ messages in thread From: Eric Biggers @ 2017-06-03 5:44 UTC (permalink / raw) To: Christoph Lameter; +Cc: akpm, viro, mm-commits, linux-fsdevel +Cc linux-fsdevel On Thu, Jun 01, 2017 at 11:07:19AM -0500, Christoph Lameter wrote: > On Wed, 31 May 2017, akpm@linux-foundation.org wrote: > > > + struct buffer_head *evictee = bh; > > + struct bh_lru *b; > > + int i; > > + b = this_cpu_ptr(&bh_lrus); > > + for (i = 0; i < BH_LRU_SIZE; i++) { > > + swap(evictee, b->bhs[i]); > > Could you try to use this_cpu_xchg here to see if it reduces latency > further? > > for (i = 0; i < BH_LRU_SIZE; i++) { > __this_cpu_xchg(bh_lrus->bhs[i], evictee) > > ... > I tried --- actually, 'evictee = __this_cpu_xchg(bh_lrus.bhs[i], evictee)'. But it's much slower, nearly as slow as the original --- which perhaps is not surprising since __this_cpu_xchg() is a cmpxchg rather than a simple load and store. It may be even worse on non-x86 architectures. Also note that we still have to disable IRQs because we need to stay on the same CPU throughout so that only a single queue is operated on. Eric ^ permalink raw reply [flat|nested] 2+ messages in thread
* Re: + fs-bufferc-make-bh_lru_install-more-efficient.patch added to -mm tree 2017-06-03 5:44 ` + fs-bufferc-make-bh_lru_install-more-efficient.patch added to -mm tree Eric Biggers @ 2017-06-05 13:51 ` Christoph Lameter 0 siblings, 0 replies; 2+ messages in thread From: Christoph Lameter @ 2017-06-05 13:51 UTC (permalink / raw) To: Eric Biggers; +Cc: akpm, viro, mm-commits, linux-fsdevel On Fri, 2 Jun 2017, Eric Biggers wrote: > I tried --- actually, 'evictee = __this_cpu_xchg(bh_lrus.bhs[i], evictee)'. But > it's much slower, nearly as slow as the original --- which perhaps is not > surprising since __this_cpu_xchg() is a cmpxchg rather than a simple load and > store. It may be even worse on non-x86 architectures. Also note that we still Its a local cmpxchg which should only take a few cycles. > have to disable IRQs because we need to stay on the same CPU throughout so that > only a single queue is operated on. Ah ok that would kill it. ^ permalink raw reply [flat|nested] 2+ messages in thread
end of thread, other threads:[~2017-06-05 13:59 UTC | newest]
Thread overview: 2+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
[not found] <592f4959.hf6G/WhBeHpccHb7%akpm@linux-foundation.org>
[not found] ` <alpine.DEB.2.20.1706011050590.8835@east.gentwo.org>
2017-06-03 5:44 ` + fs-bufferc-make-bh_lru_install-more-efficient.patch added to -mm tree Eric Biggers
2017-06-05 13:51 ` Christoph Lameter
This is a public inbox, see mirroring instructions for how to clone and mirror all data and code used for this inbox; as well as URLs for NNTP newsgroup(s).