From mboxrd@z Thu Jan 1 00:00:00 1970 From: Ville =?iso-8859-1?Q?Syrj=E4l=E4?= Subject: Re: [PATCH 0/3] drm/i915: Reduce fence usage for scanout Date: Tue, 28 Jan 2014 15:51:06 +0200 Message-ID: <20140128135106.GY9454@intel.com> References: <1390913516-730-1-git-send-email-ville.syrjala@linux.intel.com> <20140128131244.GQ2508@nuc-i3427.alporthouse.com> Mime-Version: 1.0 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable Return-path: Received: from mga11.intel.com (mga11.intel.com [192.55.52.93]) by gabe.freedesktop.org (Postfix) with ESMTP id DB4F810584A for ; Tue, 28 Jan 2014 05:51:10 -0800 (PST) Content-Disposition: inline In-Reply-To: <20140128131244.GQ2508@nuc-i3427.alporthouse.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: Chris Wilson , intel-gfx@lists.freedesktop.org List-Id: intel-gfx@lists.freedesktop.org On Tue, Jan 28, 2014 at 01:12:44PM +0000, Chris Wilson wrote: > On Tue, Jan 28, 2014 at 02:51:53PM +0200, ville.syrjala@linux.intel.com w= rote: > > From: Ville Syrj=E4l=E4 > > = > > After staring at the scanout vs. fences situation a bit, I decided > > to reduce our scanout fence usage a bit. > = > I feel that is a glorious overreaction to a bug in another piece of > code, but now that you've done it, I can not complain. ;-) > = > > Ville Syrj=E4l=E4 (3): > > drm/i915: Don't waste fences on sprite buffers > = > I suspect this is a little too generic and that there will be some > freaky old piece of hardware that requires a fence. I don't see a problem making such changes if and when needed. But if you prefer I could adjust the code a bit to have one pin function for primary planes (will check gen+FBC) and another for sprite/overlay planes (will only check gen). I suppose the current sprite code should work OKish for sprite B/C on gen2/3, but for the video overlay the code would need more changes. The hardware is more capable than the current code, and I'd hate not taking full advantage of it. > = > However, for the gen that currently use sprites, it should just work I > think. So which i-g-t hits tiled sprites? There are none. I've been waiting for someone to write some sprite i-g-ts. I do have my glplane thingy however which uses tiled buffers via gbm, and it works equally well w/ and w/o fences. > = > > drm/i915: Fix fence leaks if fbdev setup fails > = > If fbdev ever uses a tiled fb, I think we are in greater trouble. So > other than the leak is only a theorectical one, the patch looks sane. Hmm. Already forgot about that. Might be nice to make it tiled and enable FBC for it :) > > drm/i915: Don't waste fences for scanout when not needed > = > But it's gen2/3 that is short of fences (and we've a... I feel this is a > micro-optimisation, but now that it is done... Yeah well, this will only "help" gen4 desktop parts and VLV since everything else has FBC AFAIK. -- = Ville Syrj=E4l=E4 Intel OTC