[virglrenderer-devel] [PATCH v2 7/7] vrend: Use RGB1 swizzle for formats with "don't care" components

[Alexandros Frantzis]

On Thu, May 10, 2018 at 06:17:44PM +0300, Alexandros Frantzis wrote:
> On Thu, May 10, 2018 at 11:24:47AM +0300, Alexandros Frantzis wrote:
> > On Thu, May 10, 2018 at 06:19:34AM +1000, Dave Airlie wrote:
> > > On 9 May 2018 at 18:52, Alexandros Frantzis
> > > <alexandros.frantzis at collabora.com> wrote:
> > > > On Wed, May 09, 2018 at 03:11:31PM +1000, Dave Airlie wrote:
> > > >> This appears to cause some regressions in piglit.
> > > >>
> > > >> fbo-blit-stretch for example.
> > > >>
> > > >> Dave.
> > > >
> > > > Hi Dave,
> > > >
> > > > thanks for the heads-up. I will investigate.
> > > 
> > > I'm seeing some other piglit regressions around this series previous patches,
> > > 
> > > I'll try and narrow it down a bit further.
> > > 
> > > Dave.
> > 
> > Hi Dave,
> > 
> > I haven't looked in depth yet, but my first guess would be "[PATCH v2
> > 4/7] vrend: Properly handle out-of-bounds blits when blitting with GL"
> > [1].
> > 
> > I introduced this patch to fix some dEQP regressions I found when the
> > subsequent patches changed some formats to use swizzles and thus led to
> > the use of the GL blit path for these formats. The patch implements a
> > clip/stretch behavior similar to how glBlitFramebuffer works, but it's
> > possible that there is some error there or some cases I am not handling
> > properly and which are exercised by piglit.
> > 
> > Thanks,
> > Alexandros
> > 
> > [1] https://lists.freedesktop.org/archives/virglrenderer-devel/2018-April/000387.html
> 
> Hi,
> 
> I have found the issue and I am working on a solution. Stay tuned.
> 
> Thanks,
> Alexandros

Hi,

I have posted a solution for the piglit fbo-blit-stretch regression in:

https://lists.freedesktop.org/archives/virglrenderer-devel/2018-May/000457.html

Thanks,
Alexandros