ChanServ changed the topic of #dri-devel to: <ajax> nothing involved with X should ever be unable to find a bar
vliaskov has quit [Remote host closed the connection]
psykose has quit [Remote host closed the connection]
psykose has joined #dri-devel
<bnieuwenhuizen>
karolherbst: on AMD we have the same limitation (max 65535 workgroups in any dimension) and it is not the driver but either firmware or hardware
<karolherbst>
uhhh
<karolherbst>
that's quite small
<karolherbst>
anyway.. I'll wire up last_block so it's not all that terrible
swick has quit [Quit: WeeChat 1.6]
danvet has quit [Ping timeout: 480 seconds]
swick[m] has quit []
mvlad has quit [Remote host closed the connection]
<karolherbst>
ehhh.. lowering big grids is annoying :')
<jenatali>
Yep
<karolherbst>
but I guess the only sane way out is really to just waste another ubo
<karolherbst>
now that I use them for anything atm...
<karolherbst>
*not
<jenatali>
Or append it to the kernel args
<karolherbst>
I mean... atm it's a kernel arg
<karolherbst>
I could potentially mess with the existing input buffer and just reupload it all the time, which is fine
<karolherbst>
just would have to store the location of the offsets...
<karolherbst>
though that's fine I guess
<karolherbst>
yeah.. actually
co1umbarius has joined #dri-devel
columbarius has quit [Ping timeout: 480 seconds]
yuq825 has joined #dri-devel
lanodan_ has joined #dri-devel
nchery has quit [Quit: Leaving]
<karolherbst>
okay.. seems like patching the input buffer ain't that problematic afterall
<karolherbst>
and I guess if you hit the grid limit, perf of reuploading the entire thing might not even matter
<jenatali>
Heh, probably not
<jenatali>
Maybe I should see about merging mine... at some point
* jenatali
wishes I had more time for these kinds of things
heat has quit [Ping timeout: 480 seconds]
mattst88 has quit [Quit: leaving]
<karolherbst>
might make probably more sense to use a real bo for the kernel input as well instead of reuploading the entire thing, but it's also so small that it probably fits into the command buffer anyway...
<karolherbst>
:D
<karolherbst>
anyway.. now I only have to actually implement the lowering
ybogdano has quit [Read error: Connection reset by peer]
mattst88 has joined #dri-devel
YuGiOhJCJ has joined #dri-devel
kts has joined #dri-devel
YuGiOhJCJ has quit [Remote host closed the connection]
YuGiOhJCJ has joined #dri-devel
karolherbst_ has joined #dri-devel
cphealy_ has joined #dri-devel
karolherbst has quit [Ping timeout: 480 seconds]
cphealy has quit [Ping timeout: 480 seconds]
cphealy has joined #dri-devel
cphealy_ has quit [Ping timeout: 480 seconds]
bmodem has joined #dri-devel
carbonfiber has quit [Quit: Connection closed for inactivity]
Akari has joined #dri-devel
Daanct12 has joined #dri-devel
Daanct12 has quit [Remote host closed the connection]
Daanct12 has joined #dri-devel
hikiko has quit [Ping timeout: 480 seconds]
hikiko has joined #dri-devel
cphealy_ has joined #dri-devel
aravind has joined #dri-devel
cphealy has quit [Ping timeout: 480 seconds]
Daanct12 has quit [Remote host closed the connection]
Daanct12 has joined #dri-devel
Duke`` has joined #dri-devel
Daanct12 has quit [Remote host closed the connection]
Daanct12 has joined #dri-devel
mhenning has quit [Quit: mhenning]
danvet has joined #dri-devel
kts has quit [Quit: Leaving]
JohnnyonFlame has joined #dri-devel
Daanct12 has quit [Quit: Quitting]
Duke`` has quit [Ping timeout: 480 seconds]
OftenTimeConsuming has quit [synthon.oftc.net graviton.oftc.net]
dcz_ has joined #dri-devel
bgs has quit [Remote host closed the connection]
OftenTimeConsuming has joined #dri-devel
Company has quit [Quit: Leaving]
ppascher has quit [Quit: Gateway shutdown]
tzimmermann has joined #dri-devel
srslypascal is now known as Guest1744
srslypascal has joined #dri-devel
Guest1744 has quit [Ping timeout: 481 seconds]
pekkari has joined #dri-devel
ppascher has joined #dri-devel
junaid has joined #dri-devel
junaid has quit [Remote host closed the connection]
nchery has joined #dri-devel
chema has quit []
srslypascal has quit [Remote host closed the connection]
srslypascal has joined #dri-devel
rasterman has joined #dri-devel
tursulin has joined #dri-devel
frieder has joined #dri-devel
frieder has quit [Quit: Leaving]
frieder has joined #dri-devel
mvlad has joined #dri-devel
jkrzyszt has joined #dri-devel
swalker_ has joined #dri-devel
MajorBiscuit has joined #dri-devel
swalker_ is now known as Guest1747
swalker__ has joined #dri-devel
ngcortes has quit [Read error: Connection reset by peer]
OftenTimeConsuming has quit [Remote host closed the connection]
evadot_ has joined #dri-devel
sarahwalker has joined #dri-devel
OftenTimeConsuming has joined #dri-devel
Guest1747 has quit [Ping timeout: 480 seconds]
zamundaaa has quit [Read error: Connection reset by peer]
evadot has quit [Read error: Connection reset by peer]
Akari has quit [Read error: Connection reset by peer]
Akari has joined #dri-devel
tursulin has quit [Remote host closed the connection]
gallo has quit [Read error: Connection reset by peer]
jkrzyszt has quit [Read error: Connection reset by peer]
gallo has joined #dri-devel
jkrzyszt has joined #dri-devel
pochu has joined #dri-devel
swalker__ has quit [Ping timeout: 480 seconds]
lynxeye has joined #dri-devel
zamundaaa has joined #dri-devel
tursulin has joined #dri-devel
frieder has quit [Quit: Leaving]
frieder has joined #dri-devel
macromorgan is now known as Guest1767
djbw has quit [Read error: Connection reset by peer]
macromorgan has joined #dri-devel
ahajda_ has joined #dri-devel
Guest1767 has quit [Ping timeout: 480 seconds]
rasterman has quit [Remote host closed the connection]
aknautiy has quit []
aknautiy has joined #dri-devel
Daanct12 has joined #dri-devel
sgruszka has joined #dri-devel
aknautiy has left #dri-devel [#dri-devel]
aknautiy has joined #dri-devel
karolherbst_ is now known as karolherbst
pcercuei has joined #dri-devel
fab has joined #dri-devel
Daaanct12 has joined #dri-devel
Daanct12 has quit [Ping timeout: 480 seconds]
jfalempe_ has quit []
Daaanct12 has quit [Quit: Quitting]
Daanct12 has joined #dri-devel
Jeremy_Rand_Talos_ has quit [Remote host closed the connection]
frieder has quit [Quit: Leaving]
frieder has joined #dri-devel
Daanct12 has quit [Read error: Connection reset by peer]
JohnnyonFlame has quit [Ping timeout: 480 seconds]
Akari has quit [Ping timeout: 480 seconds]
fab has quit [Ping timeout: 480 seconds]
fab has joined #dri-devel
tobiasjakobi has joined #dri-devel
tobiasjakobi has quit []
jfalempe has joined #dri-devel
fab has quit [Quit: fab]
kov has quit [Quit: Coyote finally caught me]
kts has joined #dri-devel
YuGiOhJCJ has quit [Quit: YuGiOhJCJ]
MrCooper has quit [Quit: Leaving]
ivyl has quit [Quit: end of flowers]
MrCooper has joined #dri-devel
gio has quit []
<tzimmermann>
mlankhorst, the current PR for drm-misc-next will be the final one for this cycle?
ivyl has joined #dri-devel
Major_Biscuit has joined #dri-devel
OftenTimeConsuming has quit [Remote host closed the connection]
OftenTimeConsuming has joined #dri-devel
jfalempe_ has joined #dri-devel
jfalempe has quit [Read error: Connection reset by peer]
sgruszka has quit [Ping timeout: 480 seconds]
OftenTimeConsuming has quit [Remote host closed the connection]
OftenTimeConsuming has joined #dri-devel
MajorBiscuit has quit [Ping timeout: 480 seconds]
genpaku has quit [Remote host closed the connection]
<karolherbst>
some optimizations relied on the work group size to not exceed 16 bits, so things broke
<karolherbst>
but besides radeonsi no driver actually supports last_block
<karolherbst>
it's for the case where local size is all 0
<karolherbst>
anyway.. for drivers not supporting last_block, I could also lower this with an offseted last block, but... then I run into the offset issue again
<jenatali>
Oh I see it's a gallium concept
<karolherbst>
yeah.. though I think it maps to some hardware
<karolherbst>
in nvidia we also have a "CTA_RASTER_WIDTH_RESUME" field which I think is exactly that
<karolherbst>
but would hav eto play around with it
<jenatali>
Yeah for those weird grid sizes I just pick a terrible local size and keep all groups the same
<karolherbst>
yeah.. that's what I'm doing
<jenatali>
But maybe splitting a last group would be better
kts has quit [Quit: Leaving]
<karolherbst>
but having a width of 349526 doesn't work too well if the grid limit is 65535
<karolherbst>
yeah
<karolherbst>
it is
<jenatali>
Especially if the terrible local size means I already have to split the groups...
<danvet>
emersion, I didn't miss the doc patch we discussed yesterday? or whatever it was, I forgot already :-/
<emersion>
ah, no, i didn't do that one, because i'm not sure it's such a great idea
<emersion>
hm, ATOMIC=2 iirc
aravind has quit [Ping timeout: 480 seconds]
<jenatali>
Given that a bunch of the Mesa tree has dropped the need for adding r-b/a-b tags to commits, since part-of gets added and people can see who was involved from the MRs... is there a timeframe when people would consider adopting that for the whole tree?
fab has joined #dri-devel
iive has joined #dri-devel
pcercuei_ has joined #dri-devel
Duke`` has joined #dri-devel
pcercuei has quit [Ping timeout: 480 seconds]
tzimmermann_ has quit []
alyssa has joined #dri-devel
<alyssa>
anholt: Playing with piglit-runner, so far works great, thank you! :-D
<alyssa>
one small question: I see it spawning piles of windows in my Wayland session, any good way to make it not do that?
<alyssa>
I guess PIGLIT_PLATFORM=gbm ? or is there something better?
djbw has joined #dri-devel
<emersion>
a workaround would be to start a headless wayland compositor to make it happy
<alyssa>
emersion: I am using a wlroots compositor indeed ;)
<alyssa>
thanks :)
<daniels>
or weston -Bheadless-backend.so -Spiglit & WAYLAND_DISPLAY=piglit piglit-runner ...
<MrCooper>
or mutter --headless piglit-runner ...
<daniels>
gotta collect 'em all
<MrCooper>
hehe, actually need -- before piglit-runner if there are parameters after it
<daniels>
emersion: ooi, how often does your headless backend tick frame events?
<emersion>
daniels: configurable, but the default is 60Hz (goes idle when nothing changed)
bgs has joined #dri-devel
<daniels>
emersion: yeah nice; I discovered the other week that our test suite runs much faster if you make it tick infinitely fast :P
<emersion>
ahah
<daniels>
(any screenshot tests will trigger a no-planes repaint and do glReadPixels as it does so, but if that repaint doesn't happen for another 15ms ...)
cphealy has quit [Ping timeout: 480 seconds]
kts has quit [Ping timeout: 480 seconds]
junaid has joined #dri-devel
pcercuei has joined #dri-devel
pcercuei_ has quit [Ping timeout: 480 seconds]
alyssa has quit [Quit: Lost terminal]
sarahwalker has quit [Remote host closed the connection]
<Venemo>
MayeulC: great, thanks. I will try to re-test on my own Fiji soon
<Venemo>
MayeulC: currently my testing machine is otherwise indisposed but I think I can test sparse on Fiji with it next week
<MayeulC>
Venemo: Thank you :)
<MayeulC>
It would be a shame to prevent Fiji users to use that feature, although ideally the issue would need to be bissected. I hope you won't have any hangs
lemonzest has joined #dri-devel
<Venemo>
MayeulC: it is the strangest thing. with sparse enabled it would simply reboot during the test run
frieder has quit [Remote host closed the connection]
<Venemo>
it's an R9 Nano, I understand yours is the Fury right? so should be the same chip but in a different SKU as far as I understand
<MayeulC>
That's it. There's also the Fury X, but I think this one is just binned differently. It used to be possible to change the BIOS of some manufacturers to unlock some extra cores since they didn't blow efuses (assuming no hw defect), but I haven't tried that on mine :)
cphealy has joined #dri-devel
nchery_ has joined #dri-devel
nchery is now known as Guest1790
nchery_ is now known as nchery
<Venemo>
me neither, but I got it used, so who knows what they did to the poor thing before I got it
Major_Biscuit has quit [Ping timeout: 480 seconds]
<MayeulC>
That was also quite weird. My computer just completely locked up, and the only way to recover was to power cycle it by holding the power button or disconnecting the power supply.
heat has quit [Remote host closed the connection]
heat has joined #dri-devel
<MayeulC>
(I also had those issues in a few other games, I recall Rise: Son of rome, and No Man's Sky out of my head, so I started thinking that was a hw issue)
junaid has quit [Ping timeout: 480 seconds]
<Venemo>
that's not really weird, it's a normal GPU hang
<Venemo>
I recommend filing a new issue report about that. if it can be consistently reproduced, should be easy to fix
<Venemo>
if it's "random" then that can be very difficult
<MayeulC>
Oh, I thought GPU hangs would not prevent the CPU from writing logs, or prevent me from sshing in.
<MayeulC>
I could pretty consistently reproduce it, but I was unable to collect any logs or apitrace
<Venemo>
it depends. sometimes the kernel hangs too
<MayeulC>
Hmm, then how do you go about debugging that? Collect a trace on another device and replay it on the faulty one?
<Venemo>
it can happen with any driver for any piece of HW. few yrs ago I had a wifi driver which would freeze the machine completely when the AP sent a bandwidth change packet or something like that
<Venemo>
I recommend to try the usual steps we recommend for hang debugging.
<Venemo>
tho, if the machine freezes completely that may be ineffective
alanc has quit [Remote host closed the connection]
alanc has joined #dri-devel
<MayeulC>
That wifi issue sounds like a serious DOS vector... I tried setting multiple environment variables for hang detection, but it didn't prevent freezes.
<Venemo>
it won't prevent them, it will help diagnose them
<Venemo>
if you can't get any logs at all, then you can still try serial port based debugging, ie. you can connect a serial port and have the kernel send out the log on that. it's somewhat tedious but that's how I realized that the wifi driver was to blame
<MayeulC>
Well, I had nothing got persisted to disk either, so no logs, and no blinking lights either, unlike with regular panics
<Venemo>
ok, that's pretty bad indeed
<Venemo>
I can test the game for you on my Fiji if it's not too difficult to obtain that game
<Venemo>
but probably only next week or later
<Venemo>
I also don't have the serial ports at hand currently.
<MayeulC>
I should try with serial, but given the absence of blinking light I thought my chances were pretty low.
<MayeulC>
Once finish writing my PhD, I should have more time and I'll try to find a reproducer easier than "40 minutes ingame, if by chance the seed loads the problematic zone". RISE used to be easier (I had a problematic save), but I need to check if it still hangs.
rmckeever has joined #dri-devel
<Venemo>
"40 minutes ingame" sounds like a random one
bmodem has quit [Ping timeout: 480 seconds]
lynxeye has quit [Quit: Leaving.]
junaid_ has quit []
Akari has quit [Ping timeout: 480 seconds]
vliaskov has quit [Remote host closed the connection]
gio has joined #dri-devel
heat_ has joined #dri-devel
heat has quit [Read error: No route to host]
junaid has joined #dri-devel
heat_ has quit [Ping timeout: 480 seconds]
tursulin has quit [Ping timeout: 480 seconds]
jkrzyszt has quit [Ping timeout: 480 seconds]
junaid has quit [Remote host closed the connection]
hch12907 has quit []
fxkamd has quit []
fxkamd has joined #dri-devel
<anholt>
DavidHeidelberg[m]: so, ci_run_n_monitor.py doesn't seem to start jobs unless the pipeline is in a merge request. any idea what's up with that?
<DavidHeidelberg[m]>
anholt: that's weird. I using it without merge requests. It just dots ....... ?
<anholt>
I get "Detected job dependencies: a618_vk 1/9: debian/x86_build-base debian/x86_build debian/arm_build kernel+rootfs_arm64 debian-arm64" then " Duration of script execution: 0.1 minutes"
<daniels>
usually people complain that CI is too slow
mhenning has joined #dri-devel
* anholt
starts jobs manuall
<DavidHeidelberg[m]>
anholt: the job is defined correctly, maybe random failure? (Retry pipeline in top right corner and run script again?)
<anholt>
I had mistyped a job name first so everything was cancelled, perhaps the script doesn't recover starting of cancelled jobs?
<DavidHeidelberg[m]>
anholt: yeah, then it fails
<DavidHeidelberg[m]>
You have to Cancel the pipeline and retry it. The script cannot recover.
<daniels>
DavidHeidelberg[m]: can we make it retry cancelled jobs?
<anholt>
why should manual intervention be necessary?
<DavidHeidelberg[m]>
if you make typo, it'll break the Gitlab CI a bit, because it disable all jobs
<DavidHeidelberg[m]>
then if you enable them manually, they don't hold dependencies
<DavidHeidelberg[m]>
which leads into Gitlab starting jobs, which doesn't have needed dependencies finished
<DavidHeidelberg[m]>
so only solution would be implement something like --reset-pipeline option, which would do "Cancel pipeline" and "Retry" button ''press''..
samuelig has quit [Quit: Bye!]
samuelig has joined #dri-devel
samuelig has quit []
<jekstrand>
How afraid of vectors should I be for SSA RA?
<bnieuwenhuizen>
decently afraid
<jekstrand>
Ok
<bnieuwenhuizen>
cwabbott might have some good stuff though
<bnieuwenhuizen>
basically a bunch of heuristics around creating split/merge affinity group like things
samuelig has joined #dri-devel
<daniels>
DavidHeidelberg[m]: can we make it print an error when the jobs don’t start?
<mhenning>
(I don't know if that's what people actually use though or if they just split live ranges)
pekkari has quit [Quit: Konversation terminated!]
samuelig_ has joined #dri-devel
samuelig_ has quit [Remote host closed the connection]
samuelig_ has joined #dri-devel
<cwabbott>
from a quick scan, aco and ir3 both use a bit more advanced version of the tree-scan algorithm in that paper
<cwabbott>
rather than forcing it back to the "global color" at the end of the basic block, we record the variables with a color other than the global color and insert fixups only when colors differ along an edge
samuelig_ has quit []
jkrzyszt has quit [Ping timeout: 480 seconds]
<cwabbott>
(aco keeps SSA around after RA and inserts phi nodes, for ir3 I decided to insert parallel copies for phi nodes before RA and not stay in SSA afterwards so it just adds to the parallel copies)
<cwabbott>
that gets tricky when there are loops involved
samuelig_ has joined #dri-devel
samuelig_ has quit []
gouchi has joined #dri-devel
<DavidHeidelberg[m]>
daniels: that's a good question. I'll look into it, maybe if the number of target jobs is empty list we can error out or something
JohnnyonFlame has joined #dri-devel
dcz_ has quit [Ping timeout: 480 seconds]
<daniels>
DavidHeidelberg[m]: yeah, or if the jobs we try to start don’t actually start?
ubitux has joined #dri-devel
mvlad has quit [Remote host closed the connection]
<DavidHeidelberg[m]>
daniels: I tried to keep the logic simple, but I'll look what can be done to make it more robust
<Prf_Jakob>
DavidHeidelberg[m]: May I ask how you registered/identified with a matrix client on OFTC?
<DavidHeidelberg[m]>
Prf_Jakob: there is description on matrix org wiki pages ;) (sorry I'm on a phone)
<Prf_Jakob>
DavidHeidelberg[m]: Thanks, will take a look.
danilo has joined #dri-devel
<Prf_Jakob>
I was afraid of that
<Prf_Jakob>
Wronge chat :p
dakr has quit [Ping timeout: 480 seconds]
<ubitux>
hi; I switched from an intel egpu to a dedicated radeon and the colors feels off, as if a filmic curve filter was applied to them; it's somehow darker and more contrasty
<ubitux>
any idea what could cause that?
<ubitux>
it's subtle but noticeable
danilo has quit []
dakr has joined #dri-devel
<Lynne>
are you on HDMI? could be a color range issue
<Lynne>
otherwise, maybe your compositor/DE is messing with your display gamma curves
<ubitux>
I'm using DP; I have no compositor/DE, i3/X11, my setup didn't change
<Lynne>
your display decided to switch modes or reset its own curves?
<ubitux>
nah the settings were preserved afaict; it's a dumb dell 1080p, nothing fancy
blazingsun has joined #dri-devel
danvet has quit [Ping timeout: 480 seconds]
<blazingsun>
auwooo or skraito here how life been here l33t h4x0r 31337
blazingsun has quit [autokilled: Spambot. Mail support@oftc.net if you think this is in error. (2022-11-17 21:33:54)]
lemonzest has quit [Quit: WeeChat 3.6]
Haaninjo has quit [Quit: Ex-Chat]
heat_ has joined #dri-devel
fab has quit [Quit: fab]
<jenatali>
Ugh CI flakes have gotten pretty bad recently
i509VCB has joined #dri-devel
<DemiMarie>
What do OpenGL and Vulkan say about a buffer being modified while it is in use? Does it cause full-on undefined behavior (anything can happen) or will it just result in corrupted pixels on the screen?
<DemiMarie>
If it is undefined behavior, then compositors will need to perform a defensive copy of all client-supplied buffers, at a potentially substantial performance penalty.
<DemiMarie>
Malicious sandboxed clients are allowed to cause themselves to be misrendered, but not to corrupt compositor memory or interfere with the display of other clients.
nchery_ has joined #dri-devel
nchery is now known as Guest1810
nchery_ is now known as nchery
<DemiMarie>
On the other hand, if Mesa can guarantee that a client that does this will only hurt itself, then the copy can be avoided. However, this guarantee must apply to all backends, including llvmpipe/lavapipe.
ids1024[m] has joined #dri-devel
Guest1810 has quit [Ping timeout: 480 seconds]
samuelig__ has joined #dri-devel
gouchi has quit [Remote host closed the connection]
samuelig__ has quit []
samuelig__ has joined #dri-devel
alyssa has joined #dri-devel
<alyssa>
Implementing batch tracking on Asahi
<alyssa>
enabling gallium_hud is causing quake3 to flip upside down
<alyssa>
such a classic
Prf_Jakob is now known as Guest1813
Wallbraker[m] is now known as Prf_Jakob
Prf_Jakob is now known as Wallbraker
Duke`` has quit [Ping timeout: 480 seconds]
Guest1813 is now known as Prf_Jakob
pcercuei has quit [Quit: dodo]
bgs has quit [Remote host closed the connection]
Leopold_ has quit [Remote host closed the connection]
Leopold_ has joined #dri-devel
sarnex has joined #dri-devel
<anholt>
ugh. sway refuses to run as root. guess dump it for CI and try using weston headless.
Cyrinux has quit []
Cyrinux has joined #dri-devel
<emersion>
anholt: yeah, artifact from the suid-root days. next sway release will accept