On Tue, Apr 23, 2019 at 6:30 PM Eric Anholt <e...@anholt.net> wrote: > Marek Olšák <mar...@gmail.com> writes: > > > On Tue, Apr 23, 2019 at 4:39 PM Mathias Fröhlich < > mathias.froehl...@gmx.net> > > wrote: > > > >> Hi, > >> > >> On Tuesday, 23 April 2019 22:23:45 CEST Marek Olšák wrote: > >> > On Tue, Apr 23, 2019 at 4:05 PM Mathias Fröhlich < > >> mathias.froehl...@gmx.net> > >> > wrote: > >> > > >> > > Hi Marek, > >> > > > >> > > On Tuesday, 23 April 2019 20:22:15 CEST Marek Olšák wrote: > >> > > > I'd like to remove swrast from devices. It doesn't work > >> (eglInitialize > >> > > > fails) and I don't think I like swrast there. Any objections? > >> > > > >> > > Yes, how do you guarantee that at least one device can be returned > >> > > in any case? Even if no drm device is found? > >> > > The egl device query extension guarantees that there is at least one > >> > > device available. > >> > > Beside that swrast may make sense for itself, that is the reason > >> > > the swrast device is there. > >> > > > >> > > >> > If no device can be returned, the extension won't be exposed. > >> Yes, I was triggering this solution back in the days. There was (is?) no > >> infrastructure to enable > >> or disable an egl extension just past initialization or at runtime. > >> Finally Emil took the route to add swrast. > >> > >> The longer I saw the swrast device the better It appeared to me. > >> ... from the users point of view. > >> > >> > OR > >> > If there is at least 1 drm device, swrast won't be in the list. > >> > > >> > From my perspective, radeonsi is always in the list and I wouldn't > like > >> > swrast to be in the list if radeonsi is in the list. > >> > >> I see - at least I think. But really the swrast device is good as is. I > >> see plenty use cases for that. > >> Having that only present when there is no hardware device makes it less > >> useful. > >> > > > > How is swrast good? How is swrast useful if you have a supported GPU? > > EGL has the possibility of finally displacing OSMesa if we can expose > swrast devices. People doing regression testing of higher-level > software stacks (xserver, servo, etc.) need a consistent rasterizer they > can use regardless of the hardware a developer might happen to have on > their desktop. >
Then I think a separate EGL extension that exposes swrast would be better. The thing with the device extensions is that swrast is not a device. Marek
_______________________________________________ mesa-dev mailing list mesa-dev@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/mesa-dev
