[Intel-gfx] [PATCH 53/53] !UPSTREAM: drm/i915: Use MMIO flips

Mateo Lozano, Oscar oscar.mateo at intel.com
Thu Jun 19 12:33:59 CEST 2014


> -----Original Message-----
> From: Chris Wilson [mailto:chris at chris-wilson.co.uk]
> Sent: Thursday, June 19, 2014 11:14 AM
> To: Mateo Lozano, Oscar
> Cc: Daniel Vetter; intel-gfx at lists.freedesktop.org
> Subject: Re: [Intel-gfx] [PATCH 53/53] !UPSTREAM: drm/i915: Use MMIO
> flips
> 
> On Thu, Jun 19, 2014 at 09:50:33AM +0000, Mateo Lozano, Oscar wrote:
> > > -----Original Message-----
> > > From: Daniel Vetter [mailto:daniel.vetter at ffwll.ch] On Behalf Of
> > > Daniel Vetter
> > > Sent: Wednesday, June 18, 2014 10:01 PM
> > > To: Mateo Lozano, Oscar
> > > Cc: intel-gfx at lists.freedesktop.org
> > > Subject: Re: [Intel-gfx] [PATCH 53/53] !UPSTREAM: drm/i915: Use MMIO
> > > flips
> > >
> > > On Fri, Jun 13, 2014 at 04:38:11PM +0100, oscar.mateo at intel.com
> wrote:
> > > > From: Sourab Gupta <sourab.gupta at intel.com>
> > > >
> > > > If we want flips to work, either we create an Execlists-aware
> > > > version of intel_gen7_queue_flip, or we don't place commands
> > > > directly in the ringbuffer.
> > > >
> > > > When upstreamed, this patch should implement the second option:
> > >
> > > Usually we just mention such requirements in the cover letter of the
> > > series and don't include them.
> > > -Daniel
> >
> > Well, since we have it here, there is something I wanted to ask about this:
> >
> > MODULE_PARM_DESC(use_mmio_flip, "use MMIO flips (-1=never, 0=driver
> > discretion [default], 1=always)");
> >
> > The "driver discretion" option will blow up with Execlists :( How do I
> > tackle this? an execlists version of intel_gen7_queue_flip? sanitizing
> use_mmio_flip when execlists get enabled? I really don´t have a preference...
> 
> Why does execlist make it hard to add a command to the last lrc that the
> object was on?
> -Chris

Well, it´s not hard: up to v2 of the series I reused intel_ring_begin, intel_ring_emit, etc... to simply accept a context (it this case, ring->default_context) but since now we have a different set of functions, it needs some more frobbing. Now, I don´t mind doing that as a especial case, but I was told I shouldn´t worry about flips because we were changing them for MMIO flips anyway...



More information about the Intel-gfx mailing list