austriancoder changed the topic of #etnaviv to: #etnaviv - the home of the reverse-engineered Vivante GPU driver - Logs https://freenode.irclog.whitequark.org/etnaviv
JohnnyonFlame has joined #etnaviv
rhyskidd has quit [Remote host closed the connection]
rhyskidd has joined #etnaviv
pcercuei has quit [Quit: dodo]
sravn has quit [Ping timeout: 256 seconds]
sravn has joined #etnaviv
rhyskidd has quit [Remote host closed the connection]
rhyskidd has joined #etnaviv
lynxeye has joined #etnaviv
rhyskidd has quit [Quit: rhyskidd]
rhyskidd has joined #etnaviv
pcercuei has joined #etnaviv
juanrubio_ has joined #etnaviv
tlwoerner has quit [Ping timeout: 256 seconds]
berton has joined #etnaviv
tlwoerner has joined #etnaviv
cphealy has quit [Remote host closed the connection]
nathanhi- has joined #etnaviv
shoragan has quit [*.net *.split]
nathanhi has quit [*.net *.split]
shoragan has joined #etnaviv
shoragan_ has joined #etnaviv
shoragan has quit [Ping timeout: 256 seconds]
philn_ has joined #etnaviv
philn has quit [Ping timeout: 260 seconds]
cphealy has joined #etnaviv
lfa has quit [Ping timeout: 260 seconds]
<mntmn> as predicted, same early_z fix fixes KiCAD rendering on GC7000L https://twitter.com/mntmn/status/1270760872542179328?s=20
<mntmn> (i'm currently testing how stable my first patch is before publishing a draft)
<austriancoder> mntmn: please run deqp without and with your patch
<mntmn> ok
<lynxeye> mntmn: this only helps when you force-disable early_z, right? I think we are still missing the code to disable early_z when a shader uses discard.
<mntmn> lynxeye: correct. as far as i understand there is a uses_discard flag in nir structures that can be used for this, but i'm not sure where the right place for that is (in etnaviv_zsa.c? emit?)
<lynxeye> mntmn: this needs to be a derived state, as we only know at draw time if the zsa state potentially allows using early-z and the currently bound fragment shader doesn't change the fragment depth
<lynxeye> so really one of two options: a) do it directly in emit, if the logic is simple enough, or b) add a new function to derive the state to etna_state_updates[]
<mntmn> ok, thanks!
pcercuei has quit [Quit: brb]
pcercuei has joined #etnaviv
pcercuei has quit [Client Quit]
nathanhi- is now known as nathanhi
lynxeye has quit [Quit: lynxeye]
pcercuei has joined #etnaviv
tlwoerner has quit [Quit: Leaving]
berton has quit [Remote host closed the connection]
Chewi has quit [Ping timeout: 256 seconds]
Chewi has joined #etnaviv
<mntmn> the remaining xwayland glitches with GC7000L (sometimes GUI elements randomly missing/flashing) stem from this function https://gitlab.freedesktop.org/xorg/xserver/-/blob/master/glamor/glamor_render.c#L1110
<mntmn> if it's disabled (software fallback), everything looks fine
<mntmn> it looks like the render operations sometimes don't complete / get skipped
<mntmn> this also fixes the GIMP