alyssa has quit [Quit: leaving]
mkurz has joined #asahi-gpu
mkurz_ has quit [Ping timeout: 480 seconds]
cylm_ has joined #asahi-gpu
wintp has joined #asahi-gpu
cylm has joined #asahi-gpu
cylm_ has quit [Ping timeout: 480 seconds]
c10l has quit [Ping timeout: 480 seconds]
Mary6 has joined #asahi-gpu
Hibyehello has joined #asahi-gpu
Hibyehello has quit [Ping timeout: 480 seconds]
alyssa has joined #asahi-gpu
<
alyssa>
lina: I can reproduce magenta rectangles on the loading screen of supertuxkart with the current branch maybe 50% of the time
<
alyssa>
(50% of the opening of stk)
<
alyssa>
the gles3 renderer (the default on next)
<
alyssa>
this is a regression since sync
<
alyssa>
hopefully shares a root cause as some of other flakes and hopefully easier to reproduce
<
alyssa>
(IDK how to trigger the hangs yet, I just know they happen after soaktesting a few hours.....)
<
alyssa>
(if we can't sort them out we're blocked on shipping the sync stuff which is not great..)
<
alyssa>
maybe you have better tooling for debugging that sort of thing, idk
<
alyssa>
I think it's the same mailbox stupidity
<
alyssa>
manifesting as a system hang instead of a crash because release build
<
alyssa>
istg if all of our evil bugs of the past few days have just been mailbox stupidity
<
alyssa>
bumped to 256, let's see how this does
<
alyssa>
that doesn't sound great
<
alyssa>
/home/alyssa/mesa/build/src/gallium/targets/dri/apple_dri.so: agx_finish_batch_occlusion_queries+0x856e83: ../src/gallium/drivers/asahi/agx_query.c:178
<
alyssa>
/home/alyssa/mesa/build/src/gallium/targets/dri/apple_dri.so: agx_batch_cleanup+0x84eba3: ../src/gallium/drivers/asahi/agx_batch.c:240
<
alyssa>
/home/alyssa/mesa/build/src/gallium/targets/dri/apple_dri.so: agx_sync_all+0x85043f: ../src/gallium/drivers/asahi/agx_batch.c:759
<
alyssa>
glretrace: ../src/gallium/drivers/asahi/agx_query.c:178: agx_finish_batch_occlusion_queries: Assertion `query->writer == batch' failed.
<
alyssa>
that is not an assertion that should ever be failing
<
alyssa>
0xfffeb8875900, 0xaaaace571ff0
* alyssa
doesn't know what userspace VAs look like anymore
<
alyssa>
this is notably happening from the sync_all when destroying contexts so idk what else is going weird
<
alyssa>
(not a fan of that sync but that's a different issue :P)
Hibyehello has joined #asahi-gpu
<
alyssa>
I smell memory corruption memory
kilolima is now known as kl
<
alyssa>
asan is being uncooperative ;\
possiblemeatball has quit [Quit: Quit]
Hibyehello has quit [Ping timeout: 480 seconds]
<
alyssa>
../src/gallium/drivers/asahi/agx_fence.c:44:8: runtime error: member access within null pointer of type 'struct pipe_fence_handle'
<
alyssa>
probably unrelated but ubsan error
Hibyehello has joined #asahi-gpu
wintp has joined #asahi-gpu
tertu2 has quit [Ping timeout: 480 seconds]
tertu has joined #asahi-gpu
bisko has quit [Quit: My Mac has gone to sleep. ZZZzzz…]
cr1901_ has joined #asahi-gpu
cr1901 has quit [Read error: Connection reset by peer]
MajorBiscuit has joined #asahi-gpu
bisko has joined #asahi-gpu
Major_Biscuit has joined #asahi-gpu
MajorBiscuit has quit [Ping timeout: 480 seconds]
nyilas has joined #asahi-gpu
cylm has quit [Ping timeout: 480 seconds]
Misthios has quit [Quit: Misthios]
Misthios has joined #asahi-gpu
nyilas has quit [Remote host closed the connection]
kesslerd has quit [Remote host closed the connection]
kesslerd has joined #asahi-gpu
kesslerd_ has joined #asahi-gpu
cr1901_ is now known as cr1901
Major_Biscuit has quit [Ping timeout: 480 seconds]
Major_Biscuit has joined #asahi-gpu
possiblemeatball has joined #asahi-gpu
MajorBiscuit has joined #asahi-gpu
Major_Biscuit has quit [Ping timeout: 480 seconds]
nate8 has joined #asahi-gpu
hightower2 has joined #asahi-gpu
possiblemeatball has quit [Quit: Quit]
louisca has joined #asahi-gpu
louisca has quit [Remote host closed the connection]
djorz has quit [Ping timeout: 480 seconds]
hightower2 has quit [Ping timeout: 480 seconds]
Mary6 has joined #asahi-gpu
possiblemeatball has joined #asahi-gpu
Mary has joined #asahi-gpu
MajorBiscuit has quit [Quit: WeeChat 3.6]
hightower2 has joined #asahi-gpu
nate8 has quit [Quit: Leaving]
louisca has joined #asahi-gpu
louisca has quit [Remote host closed the connection]
louisca has joined #asahi-gpu
louisca has quit [Ping timeout: 480 seconds]
WindowPa- has joined #asahi-gpu
WindowPain has quit [Ping timeout: 480 seconds]
WindowPain has joined #asahi-gpu
WindowPa- has quit [Ping timeout: 480 seconds]
possiblemeatball has quit [Quit: Quit]
kesslerd_ has quit [Ping timeout: 480 seconds]
djorz has joined #asahi-gpu
c10l has joined #asahi-gpu
c10l has quit [Quit: Bye o/]
c10l has joined #asahi-gpu
<
alyssa>
lina: Worth noting the magenta rectangles also happen sporadically in other apps, but easiest to reproduce with stk
<
alyssa>
(i.e. this is blocking)
<
alyssa>
HOWEVER they don't seem to happen with ASAHI_MESA_DEBUG=sync
<
alyssa>
also can you please open MRs for your code instead of me needing to randomly cherrypick stuff
<
alyssa>
and review the MRs that are up and
<
alyssa>
uh agx/next is massive right now
<
alyssa>
actually let me land some of your stuff now
Misthios has quit [Quit: Misthios]
Misthios has joined #asahi-gpu
<
alyssa>
"Harden aganst invalid framebuffers" is totally broken for compute, dropping the commit
<
alyssa>
Failed to allocate BO VMA
<
alyssa>
sounds like we introduced a leak
<
alyssa>
er, no, just address space exhaustion
<
alyssa>
because of the new guard page stuff
<
alyssa>
32-bits of VA for USC heap
<
alyssa>
with 32*16384 bytes of guard padding after every BO, we get at most 8192 BOs
<
alyssa>
which shader-db runs out of
<
alyssa>
(I have lots of shaders)
<
alyssa>
dropping the guard page commit because I don't think it should be there by default
c10l has quit [Quit: Bye o/]
<
alyssa>
also, please review the open asahi mr's
<
alyssa>
the size of agx/next is overwhelming
cr1901_ has joined #asahi-gpu
cr1901 has quit [Read error: Connection reset by peer]
cr1901_ is now known as cr1901
c10l has joined #asahi-gpu
WindowPain has joined #asahi-gpu
c10l has joined #asahi-gpu