<
anholt>
someone was complaining that zink wasn't doing descriptor caching and descriptor overhead was high
<
zmike>
I just deleted it
<
anholt>
(on this test, where actually cpu overhead doesn't matter unless you disable submits)
<
zmike>
sounds like the angle tests I've been poking at
<
anholt>
what are you up to with angle?
<
zmike>
nothing, thankfully
myon98 has joined #zink
Sachiel has quit [Server closed connection]
Sachiel has joined #zink
cheako has joined #zink
LexSfX has quit [Remote host closed the connection]
LexSfX has joined #zink
zmike has quit [Server closed connection]
zmike has joined #zink
Simonx22 has quit [Server closed connection]
Simonx22 has joined #zink
fahien has joined #zink
FernandoMonteiro[m] has joined #zink
fahien has quit [Ping timeout: 480 seconds]
fahien has joined #zink
fahien has quit [Quit: fahien]
hch12907 has joined #zink
JulianGro[m] has joined #zink
kusma has joined #zink
neobrain[m] has joined #zink
Soroush has joined #zink
fahien has joined #zink
fahien has quit [Ping timeout: 480 seconds]
fahien has joined #zink
FernandoMonteiro[m] has quit []
neobrain[m] has quit []
fahien has quit [Ping timeout: 480 seconds]
<
zmike>
anholt: were there any other gfxbench cases remaining, or was that the last one?
<
anholt>
aztec ruins is the one that represents real rendering, and zink is at 12fps to freedreno's 16.5ish.
<
anholt>
(turnip's at 15.5ish)
<
anholt>
the main renderpass is like 1/3 the speed
<
zmike>
I was hoping that maybe it was somehow related to your descriptor findings
<
anholt>
I threw descriptor loads at it, no big effect