alyssa changed the topic of #panfrost to: Panfrost - FLOSS Mali Midgard & Bifrost - Logs https://oftc.irclog.whitequark.org/panfrost - <macc24> i have been here before it was popular
vstehle has quit [Ping timeout: 480 seconds]
adjtm has quit [Ping timeout: 480 seconds]
adjtm has joined #panfrost
karolherbst has quit [Remote host closed the connection]
karolherbst has joined #panfrost
karolherbst has quit []
karolherbst has joined #panfrost
atler is now known as Guest656
atler has joined #panfrost
Guest656 has quit [Ping timeout: 480 seconds]
sigmaris has quit [Ping timeout: 480 seconds]
sigmaris has joined #panfrost
jernej has quit [Remote host closed the connection]
jernej has joined #panfrost
anholt has quit [Remote host closed the connection]
anholt has joined #panfrost
floof58 has joined #panfrost
Macuser has joined #panfrost
Macuser has quit [Remote host closed the connection]
vstehle has joined #panfrost
Chocobozzz18 has joined #panfrost
Chocobozzz18 has quit [Remote host closed the connection]
wwilly has joined #panfrost
ced[m] has joined #panfrost
ced[m] has quit [Remote host closed the connection]
wwilly has quit [Ping timeout: 480 seconds]
<macc24>
icecream95: idk ill probably just mimick the bootlogo for that c o n s i s t e n c y
pawel has joined #panfrost
pawel has left #panfrost [#panfrost]
pawel has joined #panfrost
ryan29 has joined #panfrost
ryan29 has quit [Remote host closed the connection]
pawel has quit []
wwilly has joined #panfrost
adjtm has quit [Ping timeout: 480 seconds]
wwilly_ has joined #panfrost
wwilly has quit [Ping timeout: 480 seconds]
wwilly__ has joined #panfrost
wwilly_ has quit [Ping timeout: 480 seconds]
toozej02191586 has joined #panfrost
toozej02191586 has quit [Remote host closed the connection]
Katar has joined #panfrost
Katar has quit [Remote host closed the connection]
stepri01 has quit []
stepri01 has joined #panfrost
jernej_ has joined #panfrost
jernej has quit [Ping timeout: 480 seconds]
<robmur01>
macc24: is your firefox issue with v89? FWIW my Windows version here updated itself yesterday and I've seen rendering crapping out a few times too :(
<macc24>
robmur01: 88.0.1
<macc24>
AFAIK firefox on windows uses angle so that's another piece that can shit itself
<robmur01>
this is of course also an Adreno, but I assume it's not Rob's problem :)
<macc24>
>windows
spY|da has joined #panfrost
<robmur01>
Surface Pro X
<macc24>
last time i used windows for non-school related purposes was 3 years ago?
spY|da has quit [Remote host closed the connection]
<daniels>
robmur01: ah, so your laptop runs Mesa then
<robmur01>
daniels: who knows what horror is hidden within Qcom's Windows drivers?
<macc24>
daniels: what the fuck
<daniels>
Zink, with a twist
<HdkR>
From testing the prox on Windows. As far as I can tell the Windows stack is mostly made from gaffer tape
snowhawk has joined #panfrost
snowhawk has quit [Remote host closed the connection]
wwilly__ has quit [Ping timeout: 480 seconds]
wwilly has joined #panfrost
wwilly has quit [Remote host closed the connection]
bbrezillon has quit [Ping timeout: 480 seconds]
bbrezillon has joined #panfrost
myall has joined #panfrost
myall has quit [Remote host closed the connection]
Budd has joined #panfrost
Budd has quit [Remote host closed the connection]
Fat-Zer has quit [Ping timeout: 480 seconds]
TheLink1 has joined #panfrost
TheLink1 has quit [Remote host closed the connection]
Fat-Zer has joined #panfrost
vikiatty-M has joined #panfrost
vikiatty-M has quit [Remote host closed the connection]
hefee18 has joined #panfrost
hefee18 has quit [Remote host closed the connection]
<tomeu>
bbrezillon: ok, so my understanding of what you suggested a month ago is that each vkEvent would be backed up by a syncobj
<tomeu>
on gpu wait operations, we split the batch and pass that syncobj as one of the in_syncs of the second half
<tomeu>
on gpu set operations, we transfer the out_sync of the first half to the event's syncobj
<tomeu>
on cpu set we call DRM_IOCTL_SYNCOBJ_SIGNAL
<tomeu>
on cpu reset we call DRM_IOCTL_SYNCOBJ_RESET
<tomeu>
bbrezillon: but I don't know how we would do gpu reset
<macc24>
reset(gpu);
alpernebbi has joined #panfrost
jernej_ is now known as jernej
_whitelogger has joined #panfrost
<bbrezillon>
tomeu: hm, I think gpu set is also implemented with a batch split
<tomeu>
bbrezillon: ah yes, that's why I mentioned first half
<bbrezillon>
you just make sure the event points to the outsync of the job
<bbrezillon>
and I think gpu reset is just a CPU reset
<bbrezillon>
I recall we had a solution for that one too
<bbrezillon>
but I don't remember what that was
<tomeu>
guess I could always wait on a thread and reset it
<bbrezillon>
I don't like this idea
<alyssa>
I don't either
<bbrezillon>
we could tweak the kernel driver to unsignal out syncobjs, but I thought we ended up finding a solution that didn't require this sort of changes last time we discussed it
<bbrezillon>
so, if you pass the event syncobj as an outsync, the fence attached to it will be set to the job out_fence as soon at it's submitted
<bbrezillon>
and this new fence is unsignaled
<bbrezillon>
that's your reset
<bbrezillon>
*your GPU reset
<bbrezillon>
so, the trick with the current implem is to transfer the queue syncobj to the even syncobj just after submitting the job containing the vkCmdResetEvent() call
<bbrezillon>
tomeu: ^
antlarr8 has joined #panfrost
antlarr8 has quit [Remote host closed the connection]
RazielXYZ has joined #panfrost
RazielXYZ has quit [Remote host closed the connection]
<bbrezillon>
tomeu: you might have to queue a NULL job if the batch contains only a Reset or Signal call though (we don't support empty job submissions)
<tomeu>
cool, I think I see how that could work
<bbrezillon>
hm, actually that still doesn't work for a reset => the fence will be signaled when the NULL job is done
zaf has joined #panfrost
zaf has quit [Remote host closed the connection]
wanko has joined #panfrost
wanko has quit [Remote host closed the connection]
<bbrezillon>
OTOH, it doesn't make much sense to signal, then reset, then signal again the same fence in a command buffer, so I guess you can reset the fence using the RESET ioctl when the batch containing a reset is submitted
pikaqiu99 has joined #panfrost
pikaqiu99 has quit [Remote host closed the connection]
guppy has joined #panfrost
guppy has quit [Remote host closed the connection]
sasafras has quit [Remote host closed the connection]
Daanct12 has joined #panfrost
Danct12 has quit [Ping timeout: 480 seconds]
karolherbst has quit [Quit: Konversation terminated!]
karolherbst has joined #panfrost
a22248201 has joined #panfrost
a22248201 has quit [Remote host closed the connection]
pastly-antispam has joined #panfrost
<macc24>
contrary to a popular belief, i don't only do cadmium, i also sometimes play factorio
<alyssa>
i believe you
<macc24>
steam says that i played factorio for 40 hours this week
<macc24>
wait, no, last 2 weeks
<alyssa>
macc24: that's how long i played deqp this week
<macc24>
,_,
<alyssa>
you're probably better at factorio than i am at deqp, though