From mboxrd@z Thu Jan 1 00:00:00 1970 Date: Mon, 22 Aug 2005 09:24:57 -0700 (PDT) From: Christoph Lameter Subject: Re: [PATCH] Use deltas to replace atomic inc In-Reply-To: <20050822154300.GA29976@wotan.suse.de> Message-ID: References: <20050820005843.21ba4d9b.akpm@osdl.org> <20050822154300.GA29976@wotan.suse.de> MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Sender: owner-linux-mm@kvack.org Return-Path: To: Andi Kleen Cc: Hugh Dickins , Linus Torvalds , Andrew Morton , nickpiggin@yahoo.com.au, linux-mm@kvack.org List-ID: On Mon, 22 Aug 2005, Andi Kleen wrote: > > The page_table_lock is taken using a spin_trylock. Its skipped if > > contended. > Hmm - doesn't try lock cause a cache line bounce on the bus too? > I think it does. That would mean its latency is not much better > than a real spinlock (assuming it doesn't have to spin) Trylock does a cmpxchg on ia64 and thus acquires a exclusive cache line. So yes. But this is only done if there are updates pending. schedule() is not called that frequently. On bootup I see on average updates of 5-20 pages per mm_counter_catchup. -- To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to majordomo@kvack.org. For more info on Linux MM, see: http://www.linux-mm.org/ . Don't email: email@kvack.org