ChanServ changed the topic of #dri-devel to: <ajax> nothing involved with X should ever be unable to find a bar
jhli has joined #dri-devel
jhli has quit [Ping timeout: 480 seconds]
jhli_ has joined #dri-devel
slattann has joined #dri-devel
nchery has quit [Ping timeout: 480 seconds]
jhli_ has quit [Ping timeout: 480 seconds]
jewins has joined #dri-devel
co1umbarius has joined #dri-devel
columbarius has quit [Ping timeout: 480 seconds]
ngcortes has quit [Ping timeout: 480 seconds]
Lucretia has quit []
<graphitemaster>
You must always have noms available when fixing things.
<graphitemaster>
Get yourself a little personal fridge fore the office and a snack drawer, keep it loaded with noms.
<graphitemaster>
Charge the noms as a business expense, you're welcome.
idr_ has quit []
idr has joined #dri-devel
<idr>
pendingchaos, jekstrand: I updated the DP4A MR (!12142). Let me know if you're happy with the changes.
kurufu has quit []
kurufu has joined #dri-devel
kurufu has quit []
<alyssa>
graphitemaster: Wise.
kurufu has joined #dri-devel
kurufu has quit []
kurufu has joined #dri-devel
kurufu has quit []
kurufu has joined #dri-devel
JohnnyonFlame has joined #dri-devel
boistordu has joined #dri-devel
<jekstrand>
idr: I'll try to look tomorrow
boistordu_ex has quit [Ping timeout: 480 seconds]
camus1 has joined #dri-devel
camus has quit [Read error: Connection reset by peer]
cef is now known as Guest4867
cef has joined #dri-devel
Guest4867 has quit [Ping timeout: 480 seconds]
idr has quit [Ping timeout: 480 seconds]
sarnex has quit [Quit: Quit]
<mareko>
when I think of crocus, I imagine a croquette
sarnex has joined #dri-devel
<alyssa>
mareko: I imagine a croking magic frog exclaiming "Crocus pocus!"
ppascher has joined #dri-devel
JohnnyonFlame has quit [Read error: Connection reset by peer]
shashanks has joined #dri-devel
shashanks has quit [Remote host closed the connection]
shashanks has joined #dri-devel
YuGiOhJCJ has joined #dri-devel
Company has quit [Quit: Leaving]
macromorgan is now known as Guest4877
macromorgan has joined #dri-devel
Guest4877 has quit [Ping timeout: 480 seconds]
slattann has quit []
YuGiOhJCJ has quit [Remote host closed the connection]
YuGiOhJCJ has joined #dri-devel
thellstrom has quit [Remote host closed the connection]
slattann has joined #dri-devel
thellstrom has joined #dri-devel
jessica_24 has quit [Quit: Connection closed for inactivity]
mattrope has quit [Read error: Connection reset by peer]
jewins has quit [Read error: Connection reset by peer]
shashanks has quit [Read error: Connection reset by peer]
Daanct12 has quit [Quit: Quitting]
Danct12 has joined #dri-devel
thellstrom1 has joined #dri-devel
thellstrom has quit [Ping timeout: 480 seconds]
danvet has joined #dri-devel
alanc has quit [Remote host closed the connection]
thellstrom1 has quit []
alanc has joined #dri-devel
mlankhorst has joined #dri-devel
shashanks has joined #dri-devel
tobiasjakobi has joined #dri-devel
tobiasjakobi has quit [Remote host closed the connection]
<airlied>
ajax: started that in 12482, ran into I've no idea how xsync/xshmfence interactions work :-P
macromorgan is now known as Guest4885
macromorgan has joined #dri-devel
Guest4885 has quit [Read error: Connection reset by peer]
xexaxo_ has joined #dri-devel
xexaxo_ has quit [Ping timeout: 480 seconds]
xexaxo_ has joined #dri-devel
reductum has quit [Remote host closed the connection]
reductum has joined #dri-devel
mbrost has quit [Read error: Connection reset by peer]
shashanks has quit [Remote host closed the connection]
tobiasjakobi has joined #dri-devel
tobiasjakobi has quit [Remote host closed the connection]
xexaxo_ has quit [Ping timeout: 480 seconds]
rasterman has joined #dri-devel
Duke`` has joined #dri-devel
dj-death has joined #dri-devel
<airlied>
ajax: doh xwayland doesn't support MIT-SHM :-P
pcercuei has joined #dri-devel
Lucretia has joined #dri-devel
xexaxo_ has joined #dri-devel
slattann has quit []
<emersion>
airlied: xserver master does support it
<emersion>
hm, well, xserver release should support part of it
lynxeye has joined #dri-devel
pnowack has joined #dri-devel
slattann has joined #dri-devel
dj-death_ has joined #dri-devel
dj-death is now known as Guest4887
dj-death_ is now known as dj-death
Guest4887 has quit []
shfil has joined #dri-devel
Ahuj has joined #dri-devel
rasterman has quit [Quit: Gettin' stinky!]
lemonzest has joined #dri-devel
xexaxo_ has quit [Ping timeout: 480 seconds]
pochu has joined #dri-devel
hansg has joined #dri-devel
rando25902 has joined #dri-devel
pochu has quit [Ping timeout: 480 seconds]
rando25892 has quit [Ping timeout: 480 seconds]
<airlied>
emersion: yeah I noticed in master, I'll tset with xorg next week
thellstrom has joined #dri-devel
vivijim has joined #dri-devel
bcarvalho has quit [Remote host closed the connection]
bcarvalho has joined #dri-devel
rasterman has joined #dri-devel
xexaxo_ has joined #dri-devel
flacks has quit [Quit: Quitter]
flacks has joined #dri-devel
slattann has quit []
hansg has quit [Quit: Leaving]
xexaxo_ has quit [Ping timeout: 480 seconds]
dogukan has joined #dri-devel
camus1 has quit []
<Namarrgon>
is mesa3d.org/gitlab down-down or only maintenance-down?
<emersion>
it works for me
<Namarrgon>
i get 502s for both
<emersion>
hm, well
<emersion>
mesa3d.org works, gitlab doesn't
<shfil>
hm, CI is running tests with just asan? if yes, then I'd recommend turning on also undefined behavior sanitizer. it catches some lighter memory corruption, which are not covered by asan
iive has joined #dri-devel
<shfil>
for example indexes of c arrays
<shfil>
with cmake I like to use `-fsanitize=undefined,float-divide-by-zero,integer,implicit-conversion,implicit-integer-truncation,implicit-integer-arithmetic-value-change,local-bounds,nullability` with clang
YuGiOhJCJ has quit [Quit: YuGiOhJCJ]
<alyssa>
Test case 'dEQP-GLES31.functional.draw_indirect.compute_interop.large.drawelements_separate_grid_1000x1000_drawcount_5000'..
<alyssa>
CTS crashed on this poor board after 11 hours rrrrr
jewins has joined #dri-devel
xexaxo_ has joined #dri-devel
Company has joined #dri-devel
camus has joined #dri-devel
nirmoy has joined #dri-devel
Ahuj has quit [Ping timeout: 480 seconds]
hansg has joined #dri-devel
<jekstrand>
jenatali: CLOn12 is just 1.0, right?
<jekstrand>
jenatali: In particular, have we tested unions much?
<jenatali>
1.2
<jekstrand>
karolherbst: ^^
<jekstrand>
How do unions work, then?
<jenatali>
Uh... after explicit IO lowering I'm pretty sure it just ends up being reads and writes from memory
<jekstrand>
Wait.... How do they work in SPIR-V. SPIR-V doesn't have OpTypeUnion....
<jekstrand>
dj-death: ^^
<dj-death>
define the top structure N times with N the number of fields in the union ?
<dj-death>
and proper padding
<karolherbst>
jekstrand: ehhh... in a weird way
<bnieuwenhuizen>
I don't think SPIR-V has tybe based aliasing rules defined so you can always just cast
<bnieuwenhuizen>
type*
<karolherbst>
I know I looked at it and I think it was doing the right thing
<karolherbst>
inptrchains do have type arguments, so you just type while following the deref I think
heat has joined #dri-devel
tlwoerner has quit [Read error: Connection reset by peer]
<jenatali>
I really want godbolt or shader-playground to have an option for compiling CL C to SPIR-V
<jenatali>
It's so long from when I have to do it myself that it takes me forever to remember all the necessary command line tools to do it myself
xexaxo_ has quit [Ping timeout: 480 seconds]
mattrope has joined #dri-devel
<jenatali>
jekstrand: Looks like LLVM IR emits unions as a struct that only contains their largest member, and then casts it before accessing anything out of it
<alyssa>
baby don't fear the OOM reaper...
<jekstrand>
jenatali: Right... Makes sense, I guess.
nchery has joined #dri-devel
dviola has quit [Quit: WeeChat 3.2]
<alyssa>
does Gallium have a way to deal with failed allocations for driver internal things?
<imirkin>
no
<alyssa>
fun.
<imirkin>
i mean ... yes. context reset.
<imirkin>
report a lost context, enjoy?
<imirkin>
certain functions are allowed to fail
<imirkin>
which then will translate into a GL_OUT_OF_MEMORY
<imirkin>
(or appropriate error)
<alyssa>
Awoof.
<imirkin>
i mean ... what were you hoping for?
<imirkin>
gallium magically swoops in and gets you more memory?
<Kayden>
^^^. agreed. the whole idea is pretty flawed
<imirkin>
the idea of finite memory? yeah, definitely very flawed :)
<Kayden>
it's all well and good to think that you should handle errors, and sometimes you can, but...in the end, things just get propagated up the chain, sometimes all the way to the app
<Kayden>
and in the end, 99% of GL apps never handle GL errors
<imirkin>
i dunno who had that idea. not a very good one.
drawat_ has quit []
<Kayden>
this is why KHR_no_error is a thing
<Kayden>
it's kind of just a big game of pass the buck until it's someone else's fault
<Kayden>
(for not handling it)
<imirkin>
Kayden: the flip side is if you're using a GL lib for e.g. rendering your desktop, you're a bit SOL
<Kayden>
yep
<imirkin>
this is why i advocate against using glamor / modesetting ddx
<Kayden>
but also most failed allocations have more to do with running out of VMA than running out of memory
<imirkin>
well, there's other errors
<Kayden>
and if you're running out of VMA, more than likely, you have a nasty leak
<Kayden>
(though I have seen it happen in 32-bit apps) :(
<imirkin>
in 32-bit it happens a lot
<imirkin>
esp with a 2/2 split
<bnieuwenhuizen>
especially if somebody maps a couple of textures
<imirkin>
one of the nice things of running 32-bit apps on a 64-bit kernel is that they get to keep all 4G of the addrress space :)
zmike has quit [Remote host closed the connection]
ezequielg has quit [Read error: Connection reset by peer]
CosmicPenguin has quit [Remote host closed the connection]
hfink has quit [Remote host closed the connection]
seanpaul has quit [Remote host closed the connection]
CosmicPenguin has joined #dri-devel
ezequielg has joined #dri-devel
hfink has joined #dri-devel
seanpaul has joined #dri-devel
zmike has joined #dri-devel
<karolherbst>
imirkin: you'd think that helps
<imirkin>
karolherbst: it doesn't hurt.
<karolherbst>
doesn't help if you GPU has too much VRAM :D
<imirkin>
i know how to solve *that* problem...
<karolherbst>
:D
<karolherbst>
true
<karolherbst>
once I tried to test something with a 24GB VRAM GPU and the nvidia driver, turned out the crashing was due to the application being 32 bit and nvidia preallocating like 3.something GB of RAM ¯\_(ツ)_/¯
<imirkin>
you gotta spend ram to make ram?
<imirkin>
(certainly true for page tables ... heh)
<karolherbst>
no idea
<karolherbst>
probably some driver optimization
<karolherbst>
maybe it's just the VA
<karolherbst>
and they premaped 3.5GB of VRAM or so
<karolherbst>
because... "15% or so"
<karolherbst>
ahh that GPU has 32GB
<karolherbst>
so maybe it was 10%
jkrzyszt has joined #dri-devel
shfil has quit [Ping timeout: 480 seconds]
<alyssa>
imirkin: where can i download ram
<alyssa>
usually i'd download freeram.exe but i guess that doesn't work on linux
<alyssa>
# apt install ram
<alyssa>
no it's already installed
<alyssa>
# dpkg-reconfigure ram
<imirkin>
alyssa: you're looking for emm386
<ccr>
doublespace4ram
<FLHerne>
alyssa: There's that kernel thing where it can compress memory contents when context-switching
<MrCooper>
that one turned out to do nothing but enlarge the Windows page file :)
<imirkin>
"When Windows 95 was launched, it was widely reported that software for the operating system would be "memory hungry", requiring at least 4 megabytes of memory and preferably 8"
<robclark>
alyssa: zram swap ftw (although I guess it doesn't help for gpu memory too much unless your GEM bo's can be unpinned)
<robclark>
CrOS use zram swap to fit 8GB of chrome into 4GB of ram ;-)
<bnieuwenhuizen>
time to switch to HMM for zram swap?
<alyssa>
jenatali: Ah yes that's the one
<alyssa>
robclark: scary
jkrzyszt has quit [Ping timeout: 480 seconds]
dviola has joined #dri-devel
<kisak>
ugh, zram swap ... I had that deployed in many systems here and the near OOM scenario is extreme and painful.
hansg has quit [Remote host closed the connection]
<alyssa>
nya
<alyssa>
😺
<kisak>
^oops, zram swap might be different from zswap, which is what I had a bad time with
hansg has joined #dri-devel
<karolherbst>
yeah
<robclark>
yeah, it is just RAM backed swap device..
<karolherbst>
zram is the useful thing
hansg has quit [Remote host closed the connection]
hansg has joined #dri-devel
hansg has quit [Remote host closed the connection]
hansg has joined #dri-devel
gouchi has joined #dri-devel
shfil has quit [Ping timeout: 480 seconds]
SolarAquarion has quit []
mbrost has quit [Remote host closed the connection]
mbrost has joined #dri-devel
jewins has quit [Remote host closed the connection]
jewins has joined #dri-devel
SolarAquarion has joined #dri-devel
hansg has quit [Quit: Leaving]
sdutt has joined #dri-devel
lynxeye has quit [Quit: Leaving.]
slattann has joined #dri-devel
mlankhorst has quit [Ping timeout: 480 seconds]
tlwoerner has joined #dri-devel
RauLi has joined #dri-devel
sdutt has quit [Read error: Connection reset by peer]
RauLi has quit []
Lucretia has quit []
Lucretia has joined #dri-devel
shfil has joined #dri-devel
slattann has quit []
angerctl has joined #dri-devel
Namarrgon has quit [Ping timeout: 480 seconds]
shfil has quit [Remote host closed the connection]
shfil has joined #dri-devel
<imirkin>
mareko: can you glance at my final comment in https://gitlab.freedesktop.org/mesa/mesa/-/issues/5227 ? i'm thinking maybe we can just stick the low bits of the pointer somewhere in the draw info. or something like that.
<imirkin>
maybe onto the pipe_resource pointer, although that's pretty dirty...
<imirkin>
although you're getting rid of vertices_per_patch (and even if you weren't, we could reduce it in size, doesn't need a full byte)
slattann has joined #dri-devel
mlankhorst has joined #dri-devel
slattann has quit []
nirmoy has quit []
Daanct12 has joined #dri-devel
ngcortes has joined #dri-devel
Danct12 has quit [Ping timeout: 480 seconds]
<Kayden>
blargh. dri2 state tracker asking me to export buffers with no context
<Kayden>
have to spontaneously unsuballocate them but no context to do copies in
<Kayden>
I guess this is why radeonsi just makes a bonus context (sscreen->aux_context) for "dri asked me to do a thing but I can't" :/
sdutt has joined #dri-devel
sdutt has quit [Remote host closed the connection]
gouchi has quit [Remote host closed the connection]
bcarvalho has quit [Remote host closed the connection]
gouchi has joined #dri-devel
<alyssa>
panfrost-t860-traces is failing with crashes. I can't reproduce these crashes locally. what's CI doing special?
<mareko>
imirkin: I've replied there
<imirkin>
mareko: thanks
<imirkin>
yeah, dunno offhand what level of support is required for unaligned pointers
<imirkin>
that's another good thing to investigate :)
<anholt_>
I would start with pulling down CI's driver build and trying them. then making sure I've got the same winsys as CI is doing for trace replay. then just grabbing CI's rootfs and dumping the drivers in there to run in the same environment.
<alyssa>
it would help if I could trigger panfrost jobs manually without having to assign marge ....
<daniels>
alyssa: er, you can?
<alyssa>
manual jobs dont like me
<alyssa>
case of the fridays but this makes up for it
<alyssa>
3 Aug 20 khr_conformance ( 17) Conformance Test Submission Received for: Software in the Public Interest, Inc.
aswarup_ has joined #dri-devel
aswarup__ has joined #dri-devel
aswarup_ has quit []
aswarup__ has quit []
hanetzer has quit [Ping timeout: 480 seconds]
<daniels>
alyssa: how not?
heat has quit [Ping timeout: 480 seconds]
<alyssa>
daniels: I click arm_build containiner to start them, that goes through and all the arm builds happen, but the arm hardware jobs dont' go through
<alyssa>
(this was in MR context)
<alyssa>
I don't even get a gear to start the arm jobs manually (I did get such a gear to start freedreno jobs though)
<daniels>
alyssa: you're probably not firing enough dependent jobs
<daniels>
alyssa: if you click on 'job dependencies' from the pipeline tab, then hit 'show dependencies', you'll see it needs debian/x86_build and kernel+rootfs_arm64 from the container stage, then debian-arm64 from the build stage
<alyssa>
..........why does panfrost need debian/x86_build?
<HdkR>
I have nothing to do with this
<HdkR>
Can't blame me :P
<alyssa>
"because panfrost depends on lava-test:arm64 which extends and needs debian/x86_build"
<daniels>
yeah, I mean the GitLab CI job does run on an x86-64 machine
<daniels>
that all it does is run a Python script is neither here nor there
<alyssa>
Hum
<daniels>
(it was originally run on aarch64, but everyone else wanted to run their controllers on x86-64, so we flipped over to have one less bit of divergence)