Re: [Intel-xe] [PATCH 3/3] drm/xe: Kill execlist support.

[Rodrigo Vivi <rodrigo.vivi@intel.com>]

On Mon, Jul 24, 2023 at 09:43:24AM -0300, Lucas De Marchi wrote:
> On Mon, Jul 24, 2023 at 10:02:24AM +0200, Francois Dugast wrote:
> > On Fri, Jul 21, 2023 at 08:40:01PM +0000, Souza, Jose wrote:
> > > On Fri, 2023-07-21 at 16:19 -0400, Rodrigo Vivi wrote:
> > > > This submission backend is incomplete and unsupported.
> > > > It was left there only for some initial bring-up for comparison
> > > > and experiments. But most of Xe driver and its features rely
> > > > entirely on the GuC submission.
> > > >
> > > > If this code is not entirely broken yet, it is likely just a matter
> > > > of time.
> > > 
> > > My 2 cents here, I think it is worthy to keep it even broken to make sure Xe have a backend layered architecture and don't end up like i915 +
> > > execlist.
> > > Also for new platforms bring-up this might come handy when fixed(if the necessity comes someone will put the needed hours to make it functional
> > > again)...
> > > 
> > 
> > My 2 cents: I agree execlist has potential to be useful but only if it is fixed
> > and maintained in the long run, and (quoting the cover letter) "untested,
> > unvalidated, un-ci, unsupported" are all addressed.
> > 
> > Otherwise I think it is better to remove it and to keep it somewhere else as a
> > patch if needed in the future, because it is odd to keep code we know not to be
> > functional.
> 
> This discussion happened over and over. And it will always come back if
> we keep it like this. If/when someone steps up to add an execlist
> backend, this patch can be used as reference to be reverted.
> 
> 
> Acked-by: Lucas De Marchi <lucas.demarchi@intel.com>

Thanks for all the reviews and comments here.
After some offline discussions we decided to keep execlists for now,
but put extra work to make it work.

So I pushed the first 2 patches of this series and skipped this one.


> 
> Lucas De Marchi