From: Daniel Vetter Subject: Re: [PATCH 2/6] drm/i915/vlv: Added a rendering specific Hw WA 'WaReadAfterWriteHazard' Date: Fri, 21 Mar 2014 15:58:56 +0100 Message-ID: <20140321145856.GB30571@phenom.ffwll.local> References: <1390362310-15963-1-git-send-email-akash.goel@intel.com> <1390362310-15963-3-git-send-email-akash.goel@intel.com> <20140122105451.GK9454@intel.com> <20140122111151.GC28361@nuc-i3427.alporthouse.com> <1395402858.19358.30.camel@sourabgu-desktop> Mime-Version: 1.0 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable Return-path: Received: from mail-ee0-f48.google.com (mail-ee0-f48.google.com [74.125.83.48]) by gabe.freedesktop.org (Postfix) with ESMTP id 6F4796E169 for ; Fri, 21 Mar 2014 07:59:02 -0700 (PDT) Received: by mail-ee0-f48.google.com with SMTP id b57so1896379eek.35 for ; Fri, 21 Mar 2014 07:59:00 -0700 (PDT) Content-Disposition: inline In-Reply-To: <1395402858.19358.30.camel@sourabgu-desktop> List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: intel-gfx-bounces@lists.freedesktop.org Sender: "Intel-gfx" To: "Gupta, Sourab" Cc: "Goel, Akash" , "intel-gfx@lists.freedesktop.org" List-Id: intel-gfx@lists.freedesktop.org On Fri, Mar 21, 2014 at 11:53:40AM +0000, Gupta, Sourab wrote: > On Wed, 2014-01-22 at 11:11 +0000, Chris Wilson wrote: > > On Wed, Jan 22, 2014 at 12:54:51PM +0200, Ville Syrj=E4l=E4 wrote: > > > On Wed, Jan 22, 2014 at 09:15:06AM +0530, akash.goel@intel.com wrote: > > > > From: Akash Goel > > > > > > > > Added a new rendering specific Workaround 'WaReadAfterWriteHazard'. > > > > In this WA, need to add 12 MI Store Dword commands to ensure proper > > > > flush of h/w pipeline. > > > > > > > > Signed-off-by: Akash Goel > > > > --- > > > > drivers/gpu/drm/i915/intel_ringbuffer.c | 25 +++++++++++++++++++++= ++++ > > > > 1 file changed, 25 insertions(+) > > > > > > > > diff --git a/drivers/gpu/drm/i915/intel_ringbuffer.c b/drivers/gpu/= drm/i915/intel_ringbuffer.c > > > > index 133d273..e8ec536 100644 > > > > --- a/drivers/gpu/drm/i915/intel_ringbuffer.c > > > > +++ b/drivers/gpu/drm/i915/intel_ringbuffer.c > > > > @@ -2167,6 +2167,31 @@ intel_ring_flush_all_caches(struct intel_rin= g_buffer *ring) > > > > > > > > trace_i915_gem_ring_flush(ring, 0, I915_GEM_GPU_DOMAINS); > > > > > > > > + if (IS_VALLEYVIEW(ring->dev)) { > > > > + /* > > > > + * WaReadAfterWriteHazard > > > > + * Send a number of Store Data commands here to finish > > > > + * flushing hardware pipeline.This is needed in the case > > > > + * where the next workload tries reading from the same > > > > + * surface that this batch writes to. Without these Sto= reDWs, > > > > + * not all of the data will actually be flushd to the s= urface > > > > + * by the time the next batch starts reading it, possib= ly > > > > + * causing a small amount of corruption. > > > > + */ > > > > + int i; > > > > + ret =3D intel_ring_begin(ring, 4 * 12); > > > > > > BSpec says 8 is enough. Is Bspec incorrect. > > = > > No, these are just figures they plucked out of the air. Last I heard > > they were using 32... > > = > > > Also this workaround is also listed for everything SNB+. > > = > > And we already have a more effective workaround. > > -Chris > = > Can you please let us know whether this workaround patch is required or > not. If not, then how is this currently handled. We emit XY_SETUP_BLT before certain blt operations to insert a sufficiently long stall. The underlying bug this works around is that the cache controller of the cpu falls over in certain very specific rwm cycles. The official w/a is a full pipeline flush when switching between reading and writing to a surface, which has a horribly perf impact on the blitter. -Daniel -- = Daniel Vetter Software Engineer, Intel Corporation +41 (0) 79 365 57 48 - http://blog.ffwll.ch