ChanServ changed the topic of #dri-devel to: <ajax> nothing involved with X should ever be unable to find a bar
xexaxo has quit [Ping timeout: 480 seconds]
Company has quit [Ping timeout: 480 seconds]
columbarius has joined #dri-devel
co1umbarius has quit [Ping timeout: 480 seconds]
mbrost has quit [Read error: Connection reset by peer]
nsneck has quit [Remote host closed the connection]
MrCooper has quit [Ping timeout: 480 seconds]
camus has joined #dri-devel
khfeng has joined #dri-devel
slattann has joined #dri-devel
jewins has quit [Ping timeout: 480 seconds]
slattann has quit []
Lucretia has quit []
jcline_ has quit [Ping timeout: 480 seconds]
MrCooper has joined #dri-devel
sdutt has joined #dri-devel
YuGiOhJCJ has joined #dri-devel
Duke`` has joined #dri-devel
thellstrom has quit [Ping timeout: 480 seconds]
slattann has joined #dri-devel
thellstrom has joined #dri-devel
sdutt has quit [Ping timeout: 480 seconds]
thellstrom1 has joined #dri-devel
thellstrom has quit [Ping timeout: 480 seconds]
thellstrom1 has quit []
mlankhorst has joined #dri-devel
tobiasjakobi has joined #dri-devel
tobiasjakobi has quit [Remote host closed the connection]
Duke`` has quit [Ping timeout: 480 seconds]
Duke`` has joined #dri-devel
danvet has joined #dri-devel
rasterman has joined #dri-devel
<pq>
haasn, by "exact opposite" do you mean always "array of bytes" or something else? If array-of-bytes, it simply cannot represent any non-8-bit pixel formats which will only become more and more important in the future, e.g. 10 bits per channel. I suppose it will fall into the same inconsistency then as OpenGL, picking the problems of all format conventions.
thellstrom has joined #dri-devel
thellstrom has quit []
yoslin has quit [Read error: Connection reset by peer]
yoslin has joined #dri-devel
<pq>
karolherbst, ISTR complaints in the past about Xwayland breaking lwjgl because the names for video modes didn't quite match the strings that Xorg used. Or maybe it was between Xorg video drivers... anyway, yeah, they at least used to do the bad thing.
<pq>
haasn, anyway, the link you got to the pixel format guide or what was it called should be helpful. It's not only docs, but a tool you can use to translate from one format definition to another.
pcercuei has joined #dri-devel
pnowack has joined #dri-devel
jkrzyszt has joined #dri-devel
tarceri has joined #dri-devel
jkrzyszt has quit [Ping timeout: 480 seconds]
bcarvalho has joined #dri-devel
lynxeye has joined #dri-devel
JohnnyonFlame has quit [Ping timeout: 480 seconds]
Ahuj has joined #dri-devel
Ahuj has quit []
Ahuj has joined #dri-devel
frieder has joined #dri-devel
pjakobsson has quit [Read error: Connection reset by peer]
mnadrian has joined #dri-devel
pjakobsson has joined #dri-devel
pochu has joined #dri-devel
nadrian has quit [Ping timeout: 480 seconds]
mauld_ has quit []
mnadrian has quit [Ping timeout: 480 seconds]
mauld has joined #dri-devel
pjakobsson_ has joined #dri-devel
mnadrian has joined #dri-devel
pjakobsson has quit [Ping timeout: 480 seconds]
nirmoy has joined #dri-devel
Lucretia has joined #dri-devel
mnadrian has quit [Ping timeout: 480 seconds]
mauld has quit [Quit: WeeChat 3.2]
mauld has joined #dri-devel
vivijim has joined #dri-devel
mlankhorst has quit [Ping timeout: 480 seconds]
xexaxo has joined #dri-devel
mnadrian has joined #dri-devel
mlankhorst has joined #dri-devel
<hch12907_>
karolherbst: Hyrums Law strikes again! :P
<hch12907_>
"all observable behaviors of your system will be depended on by somebody"
shfil has joined #dri-devel
flacks has quit [Quit: Quitter]
flacks has joined #dri-devel
vivijim has quit [Ping timeout: 480 seconds]
YuGiOhJCJ has quit [Quit: YuGiOhJCJ]
xexaxo has quit [Ping timeout: 480 seconds]
thellstrom has joined #dri-devel
xexaxo has joined #dri-devel
thellstrom has quit [Ping timeout: 480 seconds]
<karolherbst>
pq: it seems like that there are some games shipping only the java stuff and not the linux native binaries :/
slattann has quit []
<karolherbst>
or maybe it's a very old lwjgl
xexaxo has quit [Ping timeout: 480 seconds]
<haasn>
pq: no, it's reversed in the sense that it calls "rgb" what drm calls "bgr", for both packed and non-packed formats
<haasn>
e.g. my rgb565 is DRM's bgr565, my rgb8 is DRM's bgr8
<haasn>
I like my convention because it's consistent with memory order for byte-aligned formats and it's also consistent with little endian conventions
<emersion>
if your convention does not depend on system endianness, then you won't have an equivalent for DRM_FORMAT_RGB565
<mlankhorst>
danvet: drm pull window is closed now right?
bcarvalho has quit [Remote host closed the connection]
pochu has quit [Ping timeout: 480 seconds]
heat has joined #dri-devel
xexaxo has quit [Ping timeout: 480 seconds]
<kusma>
jenatali, daniels: Seems like we have an odd regression in the D3D12 driver on CI, that's causing some merge-grief for me... Two primitive-id-restart tests has started failing on CI, but since last time they passed, I can't see any relevant or suspicious changes...
<daniels>
kusma: until we figure it out the right thing to do would just be to add it to skip/flake
<daniels>
has the container been rebuilt recently? maybe we've got a new WARP?
<kusma>
daniels: Good question... How do I check? :)
<kusma>
The test doesn't seem to be a flake; re-running produced the exact same two failures.
<daniels>
if you hit complete raw on the logs, the first couple of lines tell you which image it's using - which is the same here
<daniels>
so it's not that, must be something in mesa itself
<kusma>
Oh, OK.
<kusma>
daniels: The only change between those two commits that isn't to unrelated drivers or platforms, is the change from my MR itself... But that makes very little sense to me, as we're not checking for point_tri_clip in the D3D12 driver, and we're not using the draw-module's clipper either...
<kusma>
So, a change to mesa also doesn't seem right to me. Which leaves "it's ghosts" as the only viable option, I think.
<kusma>
So, something broke. And I have no clue what 😓
xexaxo has quit [Ping timeout: 480 seconds]
natur has joined #dri-devel
natur has quit []
agd5f has joined #dri-devel
<jenatali>
kusma: Could be something regressed it but didn't run those tests somehow?
<kusma>
jenatali: Doesn't seem like it.
<jenatali>
I'd expect anything that could regress it to run those tests before merge, but maybe not?
<kusma>
As I said, there's seemingly no relevant changes between the last passing and the first failing commits. I'm trying to re-run the CI for the last passing commit now...
<kusma>
So... Something outside of mesa has somehow regressed this.
<jenatali>
:O
khfeng has quit [Ping timeout: 480 seconds]
<HdkR>
Did the CI machine get hit with a Windows Update? ;)
<jenatali>
That's the only explanation I can think of, but I don't think we would've pushed out any updates to this version of Server for WARP or D3D that would only impact primitive restart for line loops
<jenatali>
Definitely ghosts
<jenatali>
kusma: I guess let's just update the baseline for now to unblock people. Want to do that or want me to?
<kusma>
Yeah, probably the right option. But ugh, trying to run this on Windows 11 shows that all primitive-id-restart tests fails to compile and validate the shaders...
<kusma>
(pso is NULL)
<kusma>
So it would seem like the validator has gotten stricter at some point, at least.
<jenatali>
kusma: Oh wait... this is a change from an expected fail to a pass
<jenatali>
I'm doing a bit of a manual "bisect" to see if I can find a commit where it starts failing for me, and I'm already back a week and it's still passing
<kusma>
Heh..
mbrost has joined #dri-devel
<jenatali>
Yeah I made it back to July and the test is still passing... I have no idea what happened but I'm just going to update the baseline
<jenatali>
I wonder if it's a combination of an underlying Windows Update plus a Mesa change somehow
jewins has joined #dri-devel
shfil has quit [Ping timeout: 480 seconds]
nchery has joined #dri-devel
<danvet>
mlankhorst, I'm assuming so
JohnnyonFlame has joined #dri-devel
<kusma>
jenatali: For me, there's no amount of going back in time that fixes the issue...
<kusma>
(or, introduces the failure on Win11)
<jenatali>
kusma: weird...
<kusma>
I mean, I "only" went back as far as first commit to the d3d12 driver after the upstreaming, but I doubt that makes a huge difference...
<kusma>
before that, it gets a bit tricky to bisect etc
dongwonk has joined #dri-devel
camus has joined #dri-devel
iive has joined #dri-devel
camus1 has quit [Remote host closed the connection]
<jenatali>
Guess it is an external fix then. Oh well
<kusma>
Yeah, I'll buy @daniels' "something change on the hosting end" explanation, even though I'm very much unhappy about it ;)
<pq>
haasn, there are no non-packed formats in DRM. And rgb565 cannot be mapped to bgr565 by just byteswap or changing the array-of-bytes vs. bits-of-word.
<pq>
haasn, if you byteswap rgb565, you get g3b5r5g3. Similarly for bgr565. Maybe you have a channel-order-reversal going on that is mistaken about pixel formats?
<pq>
haasn, as said, an array-of-bytes representation convention cannot describe many of the pixel formats. 8 bits per channel is really just an unlucky special case, because it is the only set of formats that allows swapping a format code for another by simple (endianess) byteswap.
mattrope has joined #dri-devel
nchery is now known as Guest4504
Guest4504 has quit [Read error: Connection reset by peer]
nchery has joined #dri-devel
<emersion>
i'd describe it as a *lucky* special case :P
<danvet>
airlied, jekstrand: könig pushed it to drm-misc-fixes
<jekstrand>
k
JohnnyonFlame has quit [Ping timeout: 480 seconds]
<MrCooper>
the order in which colour format components are written vs their order in memory / in a word is technically convention as well
Ahuj has quit [Ping timeout: 480 seconds]
frieder has quit [Remote host closed the connection]
xexaxo_ has joined #dri-devel
JohnnyonFlame has joined #dri-devel
shfil has joined #dri-devel
<alyssa>
Context A writes to resource R, but does not flush it.
xexaxo_ has quit [Ping timeout: 480 seconds]
<alyssa>
Context B does a CPU read of resource R, forcing a flush.
<alyssa>
Which context submits the work? Visible in the syncobj exposed from fence_create.
<alyssa>
Or... neither?
<alyssa>
Oh wait. fence_create is a panfrost internal thing used to implement flush(), not gallium. my bad.
<jenatali>
alyssa: Are these two contexts on the same thread? Switching contexts should flush
<jenatali>
If they're on separate threads... how did the app synchronize them?
<alyssa>
jenatali: Two contexts on two threads.
<alyssa>
and I don't know, I don't have an app in mind. Just have discovered that multithreading is very badly broken in panfrost and i'm trying to figure out how to fix it without locking literally everything
<jenatali>
Oh, that just sounds invalid then
<alyssa>
Hmm?
<zmike>
> wolfenstein: the new order has entered the chat
<alyssa>
/kick wolfenstein: the new order
<alyssa>
jenatali: can i RiiR panfrost
<alyssa>
memory safety is easy, thread safety is where it's at >_>
<robclark>
alyssa: I'm a fan of the thread safety annotations.. esp when you do threaded-context.. I do end up building with both clang and gcc (but mostly clang.. which is also what CrOS uses)
<alyssa>
robclark: Ack..
<alyssa>
My brain is turning into mush trying to understand this race.
shfil has quit [Ping timeout: 480 seconds]
<jekstrand>
LOCK ALL THE THINGS!!!
slattann has joined #dri-devel
<alyssa>
* jekstrand has entered the chat
thellstrom has quit [Remote host closed the connection]
jewins has quit [Remote host closed the connection]
jewins has joined #dri-devel
vivijim has quit [Remote host closed the connection]
vivijim has joined #dri-devel
rsripada has quit [Remote host closed the connection]
shankaru has quit [Remote host closed the connection]
mdnavare has quit [Remote host closed the connection]
MrCooper has quit [Remote host closed the connection]
mbrost has quit [Remote host closed the connection]
mdnavare has joined #dri-devel
dolphin has quit [Remote host closed the connection]
slattann has quit [Remote host closed the connection]
dolphin has joined #dri-devel
mbrost has joined #dri-devel
aswar002 has quit [Remote host closed the connection]
nchery has quit [Remote host closed the connection]
aswar002 has joined #dri-devel
unerlige has quit [Remote host closed the connection]
Ryback_ has quit [Remote host closed the connection]
Ryback_ has joined #dri-devel
pzanoni has quit [Remote host closed the connection]
ramaling_ has quit [Remote host closed the connection]
mattrope has quit [Remote host closed the connection]
pzanoni has joined #dri-devel
slattann has joined #dri-devel
ramaling has joined #dri-devel
shankaru has joined #dri-devel
mattrope has joined #dri-devel
MrCooper has joined #dri-devel
unerlige has joined #dri-devel
rsripada has joined #dri-devel
gouchi has joined #dri-devel
thellstrom has joined #dri-devel
idr has joined #dri-devel
nchery has joined #dri-devel
<idr>
So... make check fails in xmlconfig tests. :(
<idr>
Interesting... the script I was using does set vblank_mode=0 at the top, but it has been like that since at least 2018 (because that's when I started tracking it with git).
<idr>
And that matches the error.
<idr>
Was that test added recently(ish)?
<idr>
September 2020... that's... plausible.
<idr>
I can definitely change my script, but that's an unfortunate way to make a unit test fail. I'm surprised nobody encountered this before.
<alyssa>
ok, new plan -- easiest possible approach to make helgrind shut up, get the test to pass, and move on with the CTS go gogogo
<alyssa>
PSA: if your driver uses batch tracking and is not freedreno, it's broken..
xexaxo_ has joined #dri-devel
<imirkin>
solution: rename your driver to freedreno, and all will be well
<alyssa>
imirkin: Good plan 👍
<jenatali>
Is "batch tracking" an algorithmic approach or a utility library?
<alyssa>
jenatali: Approach, I guess.
<alyssa>
not flushing during set_framebuffer_state(), tracking dependencies of everything, and flushing only when needed
<jenatali>
Got it
<robclark>
jenatali: its the thing tilers need to do when apps do silly extra framebuffer switches (like at one point in time, stk would cycle thru all it's fbo's clearing them at the start of the frame)
Daanct12 has quit [Quit: Quitting]
Danct12 has joined #dri-devel
<jenatali>
That makes sense
ngcortes has joined #dri-devel
ngcortes has quit [Remote host closed the connection]
<alyssa>
oh and seqnums can overflow .. right ...
ngcortes has joined #dri-devel
<alyssa>
and i've just introduced a deadlock. ugh...
<alyssa>
helgrind is help
nchery has quit [Ping timeout: 480 seconds]
nchery has joined #dri-devel
Duke`` has quit [Ping timeout: 480 seconds]
gouchi has quit [Remote host closed the connection]
lynxeye has quit []
<mattst88>
anyone have any alternative suggestions for solving this MSVC compile error?
<mattst88>
I just dropped the parens around the initializer, which allows me to pass a struct {...} initializer directly, but I find it a bit distasteful
<mattst88>
I wish MSVC had statement expressions
<jekstrand>
mattst88: ugh
danvet has quit [Ping timeout: 480 seconds]
<mattst88>
indeed
Peste_Bubonica has quit [Ping timeout: 480 seconds]
<jekstrand>
mattst88: Best suggestion would be wrap the whole u_dynarray_append in another macro. Something like this maybe? #define SET_EXPECTED(...) do { result_type res = {{ __VA_ARGS__ }}; u_dynarray_append(&expect, result_type, res); } while (0)
<mattst88>
jekstrand: yeah, that seems reasonable
<mattst88>
I'll give that a go
<jekstrand>
You've basically built a meta-language so obviousness is already kind-of out the window
<mattst88>
heh, yeah
<jekstrand>
mattst88: Your use of overloading >> is simultaneously one of the best and worst uses of C++ operator overloading I think I've ever seen.
<mattst88>
lol :)
<mattst88>
yeah, for just about anything else I would *hate* it :)
<mattst88>
I briefly tried to overload -> so I could basically just drop in a dot graph, but C++ (reasonably) requires that overloaded -> must still return a pointer to a member field :)
<ajax>
i wish mesamatrix knew that there were gl versions before 3
<jekstrand>
mattst88: Yeah, I thought about that too. :)
<ajax>
anyone know what big-gl version you get on v3d?