00:09
dorkbutt has joined #asahi-gpu
00:46
nate8_ has joined #asahi-gpu
00:53
nate8 has quit [Ping timeout: 480 seconds]
00:54
Kaz has joined #asahi-gpu
01:48
dorkbutt has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
01:49
dorkbutt has joined #asahi-gpu
02:13
mental has joined #asahi-gpu
02:14
zalyx has quit [Quit: later alligator]
02:15
zalyx has joined #asahi-gpu
02:16
mental is now known as lugpocalypse
02:16
dorkbutt has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
02:17
dorkbutt has joined #asahi-gpu
02:18
lugpocalypse has quit [Quit: Konversation terminated!]
02:25
<
lina >
jannau: For gnome terminal you need to unset the use system color scheme thing, else the background breaks. I don't know what's up with that...
02:45
pjakobsson has joined #asahi-gpu
02:49
pjakobsson_ has quit [Ping timeout: 480 seconds]
03:23
dorkbutt has quit [Read error: Connection reset by peer]
05:36
captainmurphy has joined #asahi-gpu
05:45
captainmurphy has left #asahi-gpu [#asahi-gpu]
05:54
captainmurphy has joined #asahi-gpu
05:54
captainmurphy has left #asahi-gpu [#asahi-gpu]
06:06
cr1901_ has joined #asahi-gpu
06:06
cr1901 has quit [Read error: Connection reset by peer]
06:15
cr1901_ is now known as cr1901
06:32
WindowPain_ has joined #asahi-gpu
06:34
WindowPain has quit [Ping timeout: 480 seconds]
06:56
<
jannau >
all good, I was just testing mpv across a couple of window managers/compositors
07:52
SSJ_GZ has joined #asahi-gpu
08:01
captainmurphy has joined #asahi-gpu
08:02
captainmurphy has left #asahi-gpu [#asahi-gpu]
08:13
LinuxM2 has joined #asahi-gpu
08:20
captainmurphy has joined #asahi-gpu
08:31
SSJ_GZ has quit [Ping timeout: 480 seconds]
08:31
SSJ_GZ has joined #asahi-gpu
08:50
SSJ_GZ has quit [Ping timeout: 480 seconds]
08:56
pip[m]1 has joined #asahi-gpu
08:57
pip[m]1 has left #asahi-gpu [#asahi-gpu]
09:25
SSJ_GZ has joined #asahi-gpu
09:26
captainmurphy has left #asahi-gpu [#asahi-gpu]
09:31
MajorBiscuit has joined #asahi-gpu
10:14
LinuxM1 has joined #asahi-gpu
10:21
LinuxM1 has quit [Quit: Leaving]
10:25
LinuxM2 has quit [Quit: Leaving]
13:25
dordoka has joined #asahi-gpu
13:44
dordoka has quit [Quit: Konversation terminated!]
13:45
cylm has joined #asahi-gpu
13:47
dordoka has joined #asahi-gpu
13:47
dordoka has quit []
14:00
minecrell8 has quit []
14:00
minecrell has joined #asahi-gpu
14:30
DarkShadow44 has joined #asahi-gpu
15:25
mkurz has joined #asahi-gpu
15:26
mkurz has joined #asahi-gpu
15:28
mkurz has quit [Remote host closed the connection]
15:28
mkurz has joined #asahi-gpu
15:50
bcrumb has joined #asahi-gpu
16:00
bcrumb has quit [Quit: WeeChat 3.7.1]
16:02
yuyichao_ has joined #asahi-gpu
16:03
bcrumb has joined #asahi-gpu
16:03
yuyichao has quit [Read error: Connection reset by peer]
16:05
nate8_ has left #asahi-gpu [Leaving]
16:06
LinuxM1 has joined #asahi-gpu
16:21
WindowPain_ is now known as WindowPain
16:21
bcrumb has quit [Quit: WeeChat 3.7.1]
16:23
bcrumb has joined #asahi-gpu
16:29
bcrumb has joined #asahi-gpu
16:36
LinuxM1 has quit [Quit: Leaving]
16:42
LinuxM1 has joined #asahi-gpu
17:09
systwi has quit [Ping timeout: 480 seconds]
17:09
systwi has joined #asahi-gpu
17:21
WindowPain has joined #asahi-gpu
17:25
MajorBiscuit has quit [Ping timeout: 480 seconds]
17:28
WindowPain has joined #asahi-gpu
17:31
mkurz has quit [Read error: Connection reset by peer]
17:44
LinuxM1 has quit [Quit: Leaving]
17:46
bcrumb has joined #asahi-gpu
17:48
cylm has quit [Quit: WeeChat 3.6]
17:48
mkurz has joined #asahi-gpu
17:54
WindowPain has joined #asahi-gpu
17:58
cylm has joined #asahi-gpu
18:03
WindowPain_ has joined #asahi-gpu
18:06
WindowPain_ has left #asahi-gpu [#asahi-gpu]
18:07
WindowPain has quit [Ping timeout: 480 seconds]
18:11
WindowPain has joined #asahi-gpu
18:15
bcrumb has joined #asahi-gpu
18:19
WindowPain_ has joined #asahi-gpu
18:20
WindowPain has quit [Read error: Connection reset by peer]
18:20
WindowPain_ is now known as WindowPain
18:27
LinuxM1 has joined #asahi-gpu
18:40
bcrumb has quit [Quit: WeeChat 3.7.1]
18:48
LinuxM2 has joined #asahi-gpu
18:59
LinuxM2 has quit [Quit: Leaving]
19:03
mkurz has quit [Read error: Connection reset by peer]
19:07
ariel has joined #asahi-gpu
19:07
ariel is now known as Guest1386
19:08
Guest1386 is now known as facekapow
19:09
facekapow has quit []
19:33
bcrumb has joined #asahi-gpu
19:35
bcrumb has joined #asahi-gpu
19:35
mkurz has joined #asahi-gpu
19:45
erik has joined #asahi-gpu
19:46
<
corion >
The gpu driver is ridiculously nice.
19:46
bcrumb has joined #asahi-gpu
19:50
Venemo_ is now known as Venemo
20:01
erik has quit [Remote host closed the connection]
20:44
cylm has quit [Quit: WeeChat 3.6]
20:59
LinuxM1 has quit [Quit: Leaving]
21:00
LinuxM2 has joined #asahi-gpu
21:23
c10l has quit [Quit: Bye o/]
21:34
c10l has joined #asahi-gpu
21:35
LinuxM2 has quit [Remote host closed the connection]
21:36
LinuxM2 has joined #asahi-gpu
21:56
amw has quit [Ping timeout: 480 seconds]
22:23
LinuxM2 has quit [Quit: Leaving]
22:48
mkurz has quit [Ping timeout: 480 seconds]
22:55
alyssa has joined #asahi-gpu
22:57
<
alyssa >
I think this mechanism might be busted.
22:59
<
alyssa >
One bug jumps out:
22:59
<
alyssa >
we need to add AGX_VDM_STREAM_LINK_LENGTH to space at the beginning
22:59
<
alyssa >
so that when we ensure there's room each time, there's actually room + enough for a final stream link for the
*next* draw
22:59
<
alyssa >
that assert then is wrong
23:00
<
alyssa >
We actually need to assert something stronger:
23:00
<
alyssa >
assert(batch_encoder_current + AGX_VDM_STREAM_LINK_LENGTH <= batch->encoder_end && "...")
23:00
<
alyssa >
i.e. we need to make sure we left enough room for the next draw to link to a new stream
23:00
<
alyssa >
without overflow on the link
23:01
<
alyssa >
It's unclear whether that bug can be hit in practice, but if not it's only by accident
23:02
<
alyssa >
I woudl also check whether we need any special allocation for the stream
23:02
<
alyssa >
i.e. low 32-bit perhaps?
23:02
<
alyssa >
I don't remember what macOS does
23:03
<
alyssa >
it's also possibly the XML for VDM_STREAM_LINK itself is busted though I don't think so, because otherwise we couldn't trace across stream links (and we can)
23:03
<
alyssa >
ensure_cmdbuf_has_space should probably also assert its own precondition
23:03
<
alyssa >
assert(batch_encoder_current + AGX_VDM_STREAM_LINK_LENGTH <= batch->encoder_end && "...")
23:04
<
alyssa >
this indicates that the control stream ("encoder") is allocated as regular memory. so not the 32-bit thing
23:04
<
alyssa >
but it's possible there is a heavier alignment requirement
23:05
<
alyssa >
Here we allocating with 256 alignment, but maybe we need more than that
23:06
<
alyssa >
this code path only triggers at high draw call counts
23:06
<
alyssa >
which explains how I missed it
23:07
<
alyssa >
10k fish at aquarium would probably hit it, 1k would probably not
23:07
<
alyssa >
and come to think of it I've actually hit this bug twice and not debugged in either case because I thought it was something else
23:08
<
alyssa >
once in a patch to shrink the default encoder size, such that instead of allocating 0x80000 we only allocate 64k and then we end up hitting this issue at much smaller draw call counts, once in drawoverhead
23:09
SSJ_GZ has quit [Ping timeout: 480 seconds]
23:09
<
alyssa >
anyway I'm supposed to be on break right now but I was eating dinner and it clicked what was going wrong
23:09
<
alyssa >
so there's the relevant info, if jannau or lina or anyone else wants to write the fix
23:39
thevar1able has quit [Remote host closed the connection]
23:40
thevar1able has joined #asahi-gpu
23:52
thevar1able has quit [Remote host closed the connection]
23:52
thevar1able has joined #asahi-gpu
23:53
thevar1able has quit [Remote host closed the connection]
23:53
thevar1able has joined #asahi-gpu
23:54
thevar1able has quit [Remote host closed the connection]
23:54
thevar1able has joined #asahi-gpu