From mboxrd@z Thu Jan 1 00:00:00 1970 From: Maarten Lankhorst Subject: Re: [PATCH 6/7] drm/radeon: allow asynchronous waiting on foreign fences Date: Thu, 04 Sep 2014 15:57:37 +0200 Message-ID: <54086FD1.7000304@canonical.com> References: <54084F0E.9020500@canonical.com> <54085031.1080600@canonical.com> <540852F6.9040006@vodafone.de> <5408563D.8010108@canonical.com> <54085A47.1050106@vodafone.de> <54086A5B.307@vodafone.de> Mime-Version: 1.0 Content-Type: text/plain; charset="windows-1252" Content-Transfer-Encoding: quoted-printable Return-path: Received: from youngberry.canonical.com (youngberry.canonical.com [91.189.89.112]) by gabe.freedesktop.org (Postfix) with ESMTP id CD69A6E6B7 for ; Thu, 4 Sep 2014 06:57:39 -0700 (PDT) In-Reply-To: <54086A5B.307@vodafone.de> List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: dri-devel-bounces@lists.freedesktop.org Sender: "dri-devel" To: =?windows-1252?Q?Christian_K=F6nig?= , "dri-devel@lists.freedesktop.org" List-Id: dri-devel@lists.freedesktop.org Hey, Op 04-09-14 om 15:34 schreef Christian K=F6nig: >> I need to check the docs how to do this correctly, > The docs don't really cover this case. > > For the GPU waiting on an address there is an extra document just for thi= s case which I don't have at hand right now. But IIRC it was recommended to= use the local memory of the device waiting on the semaphore. I'm just not = sure if that's for pure performance reasons to avoid accessing the bus or i= f there's a hard and unavoidable hardware reason to do so. > > For the GPU signaling case there is a special bit in the semaphore instru= ctions that you need to set if any user outside of the GPU should see the w= rite. > > In general it is explicitly supported to use semaphores for inter device = synchronization on the bus (that's what the block is made for), but it's no= t intended to be used for synchronization between the CPU and the device. S= o I'm not sure if things like cache snooping is implemented and correctly s= upported. > > Well I see the feature more like nice to have and needs a bunch of testin= g, so I would say either wait with the patch for now or make it optional to= use or something like that. You're right, it's meant as something 'nice to have'. This is why it came a= fter the patch that exports reservation_object to/from dma-buf. :-) ~Maarten