ChanServ changed the topic of #zink to: official development channel for the mesa3d zink driver || https://docs.mesa3d.org/drivers/zink.html
pendingchaos_ has joined #zink
pendingchaos has quit [Ping timeout: 480 seconds]
fahien has joined #zink
fahien has quit [Ping timeout: 480 seconds]
fahien has joined #zink
JulianGro[m] has quit [Write error: connection closed]
hch12907 has quit [Write error: connection closed]
Soroush has quit [Write error: connection closed]
kusma has quit [Write error: connection closed]
neobrain[m] has quit [Write error: connection closed]
hch12907 has joined #zink
fahien has quit [Ping timeout: 480 seconds]
fahien has joined #zink
pendingchaos_ is now known as pendingchaos
JulianGro[m] has joined #zink
kusma has joined #zink
neobrain[m] has joined #zink
Soroush has joined #zink
italove9 has quit []
italove9 has joined #zink
fahien has quit [Ping timeout: 480 seconds]
Sachiel has quit [Ping timeout: 480 seconds]
Sachiel has joined #zink
_whitelogger has joined #zink
Soroush has quit [Write error: connection closed]
JulianGro[m] has quit [Write error: connection closed]
hch12907 has quit [Write error: connection closed]
neobrain[m] has quit [Write error: connection closed]
kusma has quit [Write error: connection closed]
hch12907 has joined #zink
mceier has joined #zink
DragoonAethis has joined #zink
<anholt> am I going to have to figure out how to rig up shader-db for zink? Because I don't want to.
<zmike> I'm gonna do it next week
<zmike> or at least that was my plan
<zmike> I looked a while ago and it was impossible without precompile
<zmike> but I think it may be possible now
LexSfX has quit [Ping timeout: 480 seconds]
JulianGro[m] has joined #zink
kusma has joined #zink
neobrain[m] has joined #zink
Soroush has joined #zink
<zmike> anholt: some new flakes in this job https://gitlab.freedesktop.org/mesa/mesa/-/jobs/29410473
<anholt> looks like some oomkilling
<anholt> oof. asphalt_9 perf looks like it's probably that we slowly allocate up to 112MB of descriptor pools across ~600 separate allocations.
<zmike> there's an extension coming that will probably fix that
<anholt> angle ends up with just 18 pools, 1.6MB. they don't suballocate images, though, so they end up sucking too.
<zmike> yeah different descriptor models
<anholt> hmm, not just descriptors, though. zink's working set gets up to 850MB, to angle's 550.
<zmike> you could cut it down to fewer allocations by bumping MAX_LAZY_DESCRIPTORS or to potentially fewer unused descriptors by reducing it