From mboxrd@z Thu Jan 1 00:00:00 1970 From: Chris Wilson Subject: Re: [PATCH 10/10] drm/i915: remove "unclaimed register" checks from I915_WRITE Date: Fri, 18 Jan 2013 20:56:39 +0000 Message-ID: <84c8a8$7dmhop@orsmga001.jf.intel.com> References: <1358540953-3979-1-git-send-email-przanoni@gmail.com> <1358540953-3979-11-git-send-email-przanoni@gmail.com> <20130118204913.GA21692@lundgren.amr.corp.intel.com> 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 B6C10E5C05 for ; Fri, 18 Jan 2013 12:57:14 -0800 (PST) In-Reply-To: <20130118204913.GA21692@lundgren.amr.corp.intel.com> List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Sender: intel-gfx-bounces+gcfxdi-intel-gfx=m.gmane.org@lists.freedesktop.org Errors-To: intel-gfx-bounces+gcfxdi-intel-gfx=m.gmane.org@lists.freedesktop.org To: Ben Widawsky , Paulo Zanoni Cc: intel-gfx@lists.freedesktop.org, Paulo Zanoni List-Id: intel-gfx@lists.freedesktop.org On Fri, 18 Jan 2013 12:49:13 -0800, Ben Widawsky wrote: > On Fri, Jan 18, 2013 at 06:29:12PM -0200, Paulo Zanoni wrote: > > From: Paulo Zanoni > > > > Some developers don't really like this code polluting I915_WRITE, and > > we've never really measured its negative impacts. So now that we > > properly print ERR_INT interrupts, let's remove the I915_WRITE code > > and promote the interrupt error message to DRM_ERROR. > > > > The downside of this change is that we lose the ability to check the > > register and print nice backtraces, but at this point most of the > > errors have already been fixed and we're investigating the few > > remaining cases. > > > > Signed-off-by: Paulo Zanoni > > I'm really sad to see this go. Especially since our time between new > platform bring-up is decreasing so much. If I were to guess right, every > developer working on a new platform would want this. So while HSW may be > in the clear, HSW+1 suffers. I don't like the extra work per iowrite32, but I can live with for the error-detection. I would rather remove it for special cases where it has demonstrable impact, perhaps in execbuffer. However, you can equally lay the blame there for execbuffer hitting i915_write32 too often. So even there I am not convinced that speeding up i915_write32 is the best approach. So NAKed-by: Chris Wilson unless you can find a way to work out the faulting register address in the interrupt or that is the only method going forward. -Chris -- Chris Wilson, Intel Open Source Technology Centre