From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from mail-io0-f169.google.com (mail-io0-f169.google.com [209.85.223.169]) by kanga.kvack.org (Postfix) with ESMTP id 7021C6B0258 for ; Wed, 9 Dec 2015 05:00:17 -0500 (EST) Received: by ioir85 with SMTP id r85so53669607ioi.1 for ; Wed, 09 Dec 2015 02:00:17 -0800 (PST) Received: from emea01-db3-obe.outbound.protection.outlook.com (mail-db3on0083.outbound.protection.outlook.com. [157.55.234.83]) by mx.google.com with ESMTPS id w4si23723471igl.28.2015.12.09.02.00.16 for (version=TLS1_2 cipher=ECDHE-RSA-AES128-SHA bits=128/128); Wed, 09 Dec 2015 02:00:16 -0800 (PST) From: Shachar Raindel Subject: RE: [RFC contig pages support 1/2] IB: Supports contiguous memory operations Date: Wed, 9 Dec 2015 10:00:02 +0000 Message-ID: References: <1449587707-24214-1-git-send-email-yishaih@mellanox.com> <1449587707-24214-2-git-send-email-yishaih@mellanox.com> <20151208151852.GA6688@infradead.org> <20151208171542.GB13549@obsidianresearch.com> In-Reply-To: <20151208171542.GB13549@obsidianresearch.com> Content-Language: en-US Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: quoted-printable MIME-Version: 1.0 Sender: owner-linux-mm@kvack.org List-ID: To: Jason Gunthorpe , Christoph Hellwig Cc: Yishai Hadas , "dledford@redhat.com" , "linux-rdma@vger.kernel.org" , Or Gerlitz , Tal Alon , "linux-mm@kvack.org" > -----Original Message----- > From: owner-linux-mm@kvack.org [mailto:owner-linux-mm@kvack.org] On > Behalf Of Jason Gunthorpe > Sent: Tuesday, December 08, 2015 7:16 PM > To: Christoph Hellwig > Cc: Yishai Hadas ; dledford@redhat.com; linux- > rdma@vger.kernel.org; Or Gerlitz ; Tal Alon > ; linux-mm@kvack.org > Subject: Re: [RFC contig pages support 1/2] IB: Supports contiguous > memory operations >=20 > On Tue, Dec 08, 2015 at 07:18:52AM -0800, Christoph Hellwig wrote: > > There is absolutely nothing IB specific here. If you want to support > > anonymous mmaps to allocate large contiguous pages work with the MM > > folks on providing that in a generic fashion. >=20 > Yes please. >=20 Note that other HW vendors are developing similar solutions, see for exampl= e: http://www.slideshare.net/linaroorg/hkg15106-replacing-cmem-meeting-tis-= soc-shared-buffer-allocation-management-and-address-translation-requirement= s > We already have huge page mmaps, how much win is had by going from > huge page maps to this contiguous map? >=20 As far as gain is concerned, we are seeing gains in two cases here: 1. If the system has lots of non-fragmented, free memory, you can create la= rge contig blocks that are above the CPU huge page size. 2. If the system memory is very fragmented, you cannot allocate huge pages.= However, an API that allows you to create small (i.e. 64KB, 128KB, etc.) c= ontig blocks reduces the load on the HW page tables and caches. Thanks, --Shachar -- 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