From: "Volkin, Bradley D" Subject: Re: [PATCH 00/13] Gen7 batch buffer command parser Date: Wed, 5 Mar 2014 09:45:29 -0800 Message-ID: <20140305174529.GA28943@bdvolkin-ubuntu-desktop> References: <1392747357-25703-1-git-send-email-bradley.d.volkin@intel.com> <20140305104635.GY17001@phenom.ffwll.local> <20140305165956.GA27629@bdvolkin-ubuntu-desktop> <20140305171438.GL17001@phenom.ffwll.local> Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit Return-path: Received: from mga02.intel.com (mga02.intel.com [134.134.136.20]) by gabe.freedesktop.org (Postfix) with ESMTP id 2CC1CFA884 for ; Wed, 5 Mar 2014 09:45:29 -0800 (PST) Content-Disposition: inline In-Reply-To: <20140305171438.GL17001@phenom.ffwll.local> List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Sender: intel-gfx-bounces@lists.freedesktop.org Errors-To: intel-gfx-bounces@lists.freedesktop.org To: Daniel Vetter Cc: "intel-gfx@lists.freedesktop.org" List-Id: intel-gfx@lists.freedesktop.org On Wed, Mar 05, 2014 at 09:14:38AM -0800, Daniel Vetter wrote: > On Wed, Mar 05, 2014 at 08:59:56AM -0800, Volkin, Bradley D wrote: > > On Wed, Mar 05, 2014 at 02:46:35AM -0800, Daniel Vetter wrote: > > > On Tue, Feb 18, 2014 at 10:15:44AM -0800, bradley.d.volkin@intel.com wrote: > > > > From: Brad Volkin > > > > 3) Coherency. I've previously found a coherency issue on VLV when reading the > > > > batch buffer from the CPU during execbuffer2. Userspace writes the batch via > > > > pwrite fast path before calling execbuffer2. The parser reads stale data. > > > > This works fine on IVB and HSW, so I believe it's an LLC vs. non-LLC issue. > > > > It's possible that the shmem pread refactoring fixes this, I just have not > > > > been able to retest due to lack of a VLV system. > > > > > > Is it still true that we need to test this on vlv? The shmem_pread path > > > really should have fixed this ... Otherwise I think this looks ready to go > > > in, I'll pester Jani for the review. > > > > Yes, I still don't have a system to test on. > > Steal one? I guess we'll notice when our QA runs all this stuff on theirs > ;-) It's not stealing if I give it back :) Anyhow, sorry, the note isn't totally clear. It shouldn't be a problem for now because the parser is still disabled on VLV (and platforms with ppgtt disabled in general). I will look into retesting this though. Brad > > Cheers, Daniel > -- > Daniel Vetter > Software Engineer, Intel Corporation > +41 (0) 79 365 57 48 - http://blog.ffwll.ch