00:31
ecloud has joined #lima
00:34
ente has quit [Ping timeout: 240 seconds]
00:34
ecloud_wfh has quit [Ping timeout: 240 seconds]
00:34
ente has joined #lima
01:20
_whitelogger has joined #lima
02:05
mripard has quit [Ping timeout: 255 seconds]
02:08
mripard has joined #lima
03:21
minicom8 has joined #lima
03:28
chewitt has quit [*.net *.split]
03:28
tlwoerner has quit [*.net *.split]
03:28
enunes has quit [*.net *.split]
03:28
minicom has quit [*.net *.split]
03:36
tlwoerner has joined #lima
03:55
dddddd has quit [Ping timeout: 256 seconds]
04:21
<
anarsoul >
testing it now
04:38
_whitelogger has joined #lima
04:52
<
anarsoul >
z16 still fails
05:14
<
MoeIcenowy >
anarsoul: mask Z16?
05:14
<
MoeIcenowy >
Z16 may need another shader
05:14
<
anarsoul >
I doubt it
05:14
<
anarsoul >
I think I understand why we need separate format for Z24S8
05:15
<
anarsoul >
basically we can't sample stencil in GLES2
05:15
<
anarsoul >
so old format is actually Z24X8
05:15
<
anarsoul >
and new is Z24S8
05:15
<
MoeIcenowy >
I think when Z16, it samples Z to the X part of the vector
05:15
<
MoeIcenowy >
but Z24S8 samples Z to Z part
05:16
<
anarsoul >
actually it samples to xyz
05:16
<
anarsoul >
both should do that
05:16
<
anarsoul >
since Z16 is actually L16
05:17
<
anarsoul >
so it's (L,L,L,1.0)
05:17
<
anarsoul >
for Z it should be (Z,Z,Z,1.0)
05:26
<
MoeIcenowy >
anarsoul: is it a repeat
05:27
<
MoeIcenowy >
or is more variables used for higher accuracy?
05:27
<
anarsoul >
it's broadcast to multiple channels
05:27
<
MoeIcenowy >
anarsoul: on 0x32 format it's not broadcast
05:27
<
MoeIcenowy >
xyz has different values
05:36
<
anarsoul >
we also have a lot of failures with depth16 in deqp-lima-fails
05:53
chewitt has joined #lima
05:58
buzzmarshall has quit [Remote host closed the connection]
06:01
<
anarsoul >
rellla: why do we dump only one texture descriptor?
06:02
<
anarsoul >
we can have up to 16
06:19
<
rellla >
anarsoul: because it's not implemented?
06:20
<
anarsoul >
having up to 16 textures is implemented :)
06:21
<
anarsoul >
MoeIcenowy: tests seem to fail because we're doing unnecessary zs reload at the very beginning
06:22
<
anarsoul >
I'm debugging why we're doing it...
06:22
<
anarsoul >
(I already fixed few bugs that I introduced)
06:26
<
anarsoul >
yeah, works now :)
06:35
<
rellla >
anarsoul: iirc i already knew, that i missed sth while introducing texture dumps :p i can do that the next days...
06:50
_whitelogger has joined #lima
07:01
<
anarsoul >
rellla: that's not something urgent :)
07:04
<
anarsoul >
MoeIcenowy: so recreate_colorbuffer tests are still failing :(
07:04
<
anarsoul >
but looks like regressions are fixed
07:09
<
anarsoul >
btw these tests also fail on blob
07:09
<
anarsoul >
hm, hold on
07:10
<
anarsoul >
they look like lima fails
07:12
<
rellla >
yes they are. also the skips.
07:12
<
anarsoul >
where I can find list of mali fails?
07:13
<
rellla >
they are mali fails iirc
07:13
<
anarsoul >
why do they match lima fails? :)
07:14
<
rellla >
not sure if i'm right atm :)
07:15
<
anarsoul >
yeah, that looks better
08:14
_whitelogger has joined #lima
09:08
kaspter has quit [Remote host closed the connection]
09:08
kaspter has joined #lima
10:11
<
MoeIcenowy >
anarsoul: maybe we hit the zone of HW bug
10:28
minicom8 is now known as minicom
10:49
rak-zero has joined #lima
11:55
chewitt has quit [Ping timeout: 265 seconds]
12:42
<
MoeIcenowy >
anarsoul: do you know how can let blob use Z16?
14:06
buzzmarshall has joined #lima
14:41
chewitt has joined #lima
14:47
marcodiego has joined #lima
14:55
dddddd has joined #lima
16:40
<
anarsoul >
MoeIcenowy: no idea
16:40
<
anarsoul >
but the point is that we don't even attempt to reload zsbuf in dEQP-GLES2.functional.fbo.render.recreate_colorbuffer.no_rebind_rbo_rgb565_depth_component16
16:42
<
anarsoul >
oh, we do, my traces are wrong
17:35
<
anarsoul >
key.zsbuf->texture->format is argb
17:39
<
MoeIcenowy >
anarsoul: what the hell
17:42
<
anarsoul >
ah, sorry, I'm wrong, it's Z16
17:45
<
anarsoul >
but why is it Z16
17:45
<
anarsoul >
I commented it out
18:02
<
anarsoul >
one more place to comment it out
18:03
<
anarsoul >
well, I promoted it to Z24
18:03
<
anarsoul >
but it still doesn't work
23:00
<
anarsoul >
MoeIcenowy: so yeah, maybe we get Z24 split into r,g,b components with 0x32 format
23:06
<
anarsoul >
btw blob uses z24s8 for z16
23:07
<
anarsoul >
I guess we should do that too
23:50
<
anarsoul >
OK, depth reload now works fine with z24s8
23:50
<
anarsoul >
stencil is broken atm
23:51
<
anarsoul >
shared_depthbuffer failures are not related to missing depth/stencil reload since they don't need it
23:51
<
anarsoul >
looking into stencil
23:53
<
anarsoul >
looks like we need some extra magic in rsw :)