From mboxrd@z Thu Jan 1 00:00:00 1970 Date: Sat, 14 Jul 2007 08:07:43 -0700 (PDT) From: Christoph Lameter Subject: Re: [PATCH 0/7] Sparsemem Virtual Memmap V5 In-Reply-To: <20070714084931.GE1198@wotan.suse.de> Message-ID: References: <617E1C2C70743745A92448908E030B2A01EA6524@scsmsx411.amr.corp.intel.com> <20070714084931.GE1198@wotan.suse.de> MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Sender: owner-linux-mm@kvack.org Return-Path: To: Nick Piggin Cc: "Luck, Tony" , Andrew Morton , linux-mm@kvack.org, linux-arch@vger.kernel.org, Andy Whitcroft , Mel Gorman List-ID: On Sat, 14 Jul 2007, Nick Piggin wrote: > Isn't it still possible that you could have TLB pressure that would > result in lower performance? I wonder why the large page support for > ia64 was shelved? 16M Large memmap support was shelved because 16M is too large a size for a vmemmap block. It results in the vmemmap overlapping multiple nodes. The TLB pressure for the 16k support is the same since its the same algorithm. We are measuring discontig/vmemmap 16k against sparse/vmemmap 16k here. We would likely see a different if we would compare sparsemem vs. sparse/vmemmap. Then there may be a difference in TLB pressure. But 16k discontig/vmemmap is the current default on IA64. -- 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