Re: [PATCH 3/3] drm/radeon: add userfence IOCTL

From: Daniel Vetter <daniel@ffwll.ch>
To: Jerome Glisse <j.glisse@gmail.com>
Cc: Serguei.Sagalovitch@amd.com, dri-devel@lists.freedesktop.org
Subject: Re: [PATCH 3/3] drm/radeon: add userfence IOCTL
Date: Tue, 14 Apr 2015 10:17:12 +0200	[thread overview]
Message-ID: <20150414081712.GF6092@phenom.ffwll.local> (raw)
In-Reply-To: <20150413194851.GH29334@gmail.com>

On Mon, Apr 13, 2015 at 03:48:51PM -0400, Jerome Glisse wrote:
> On Mon, Apr 13, 2015 at 08:26:05PM +0200, Christian König wrote:
> > On 13.04.2015 19:51, Jerome Glisse wrote:
> > >On Mon, Apr 13, 2015 at 07:23:34PM +0200, Daniel Vetter wrote:
> > >>On Mon, Apr 13, 2015 at 04:52:17PM +0200, Christian König wrote:
> > >>>From: Christian König <christian.koenig@amd.com>
> > >>>
> > >>>WIP patch which adds an user fence IOCTL.
> > >>>
> > >>>Signed-off-by: Christian König <christian.koenig@amd.com>
> > >>I've discussed userspace fences a lot with Jerome last XDC, so here's my
> > >>comments:
> > >>
> > >>My primary concern with mid-batch fences is that if we create real kernel
> > >>fences (which might even escape to other places using android syncpts or
> > >>dma-buf) then we end up relying upon correct userspace to not hang the
> > >>kernel, which isn't good.
> > >Yes i agree on that, solution i propose make sure that this can not happen.
> > 
> > What if we want to base a GPU scheduler and Android sync points on that
> > functionality? E.g. it might be necessary to create "struct fence" objects
> > which are based on the information from userspace.
> > 
> > Would that be possible or would we run into issues?
> 
> Well i would not like that, but i do not like Android code much, but you
> could use the same code as i show in my other email and just have the
> android fence struct take a reference on the BO where the fence is. Then
> using same code to check status.
> 
> Obviously there should be timeout as there is no way for the kernel to
> know if such fence will ever signal.

Yeah I think if your umd is using this for all fencing needs, including
cross-process and everything then I don't think it's such a good idea any
more ;-) But if it's just fine-grained sync for heterogenous compute
within the same process (ocl, hsa or whatever) then this seems reasonable.

I guess this boils down to what the userspace side will look like, and
what kind of standard you're trying to implement here.
-Daniel
-- 
Daniel Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch
_______________________________________________
dri-devel mailing list
dri-devel@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/dri-devel