rasterman has quit []
thellstrom1 has joined #dri-devel
thellstrom has quit [Remote host closed the connection]
<FLHerne> Venemo: As of next Mesa release, all supported HW will have a Gallium driver?
<FLHerne> Iris/Crocus take away the "doesn't work on Intel" thing
<FLHerne> I suppose everyone has to support the Nvidia blob though :-(
bloom has joined #dri-devel
<bloom> Is it just me or is driver_location mostly useless?
<bloom> If set early, other passes stomp over it
<bloom> If set late (i.e. after lower_io), it doesn't propagate to the intrinsics
<bloom> jekstrand: ^
<bloom> I mean I guess it's used internally in NIR for things
<bloom> but it's hardly /driver/_location at that point
<bloom> for actual driver location I think you need an out-of-band thing ...
VanUnamed has joined #dri-devel
VanUnamed has quit [Remote host closed the connection]
vx26 has joined #dri-devel
vx26 has quit [Remote host closed the connection]
haljordan has joined #dri-devel
haljordan has quit [Remote host closed the connection]
Lightkey has quit [Ping timeout: 480 seconds]
boistordu has joined #dri-devel
Lightkey has joined #dri-devel
boistordu_ex has quit [Ping timeout: 480 seconds]
peaklight has joined #dri-devel
peaklight has left #dri-devel [#dri-devel]
<Venemo> mareko: FLHerne I meant that many DXVK users are on NVidia's proprietary driver
<Venemo> bloom: after lower_io you need to deal with the IO intrinsics individually and not the variables. What are you trying to do?
<bloom> Venemo: remap varying locations to make room for driver-internal state and packing
<bloom> which the more I think about it seems even more silly and impossible to work in light of separable shaders
<bloom> so err I think that's my cue to try to understand the hardware better ....
camus has joined #dri-devel
<Venemo> bloom: which driver are you working on?
<bloom> asahi (apple)
<Venemo> Well, if you do that before lower_io then the variables' driver location is what you need to set
<bloom> That's what I thought, but that doesn't work if you call any other NIR pass that wants to remap driver locations / bases itself
<bloom> nir_lower_mediump_io was the culprit for the bug I chased down tonight, but grepping nir it looks like there might be other passes that do similar things ... so apparently this is not something that we can expect to work for NIR
alyssa has left #dri-devel [#dri-devel]
<Venemo> Usually the passes don't really need to touch driver locations, but rather the driver has a place that sets it
<Venemo> For example, radv_set_driver_locations
<bloom> so maybe this is just a nir_lower_mediump_io bug and not a NIR wide problem?
Daanct12 has joined #dri-devel
<Venemo> What exactly is the bug? nir_lower_mediump doesn't seem to touch driver locations
<Venemo> Hm, wait looks like it does
<bloom> This function
<bloom> lower_mediump_io only makes sense to call after lower_io
<bloom> but that function recomputes all the `base` values, thereby scribbling over all the driver_locations
<Venemo> Yes, it seems to be intended to work after lower_io since it works on IO semantic locations
<bloom> and works on load_input and friends instead of derefs
<Venemo> Yeah
<Venemo> I see what you mean
Daaanct12 has quit [Ping timeout: 480 seconds]
<Venemo> Sounds like if you want to reserve locations for your own purposes then it sounds like this pass could use some improvement
<bloom> Weee.
<bloom> pass is also problematic for unrelated reasons on Mali
<bloom> mediump i/o is a hard problem 🤷
<bloom> (adds too many conversions)
<HdkR> Everyone hates supporting mediump but everyone loves the "doubling" of ALU performance :P
luzipher_ has joined #dri-devel
luzipher__ has quit [Ping timeout: 480 seconds]
drawat has quit [Ping timeout: 480 seconds]
nekron has joined #dri-devel
nekron has quit [Remote host closed the connection]
via19 has joined #dri-devel
via19 has quit [Remote host closed the connection]
drawat has joined #dri-devel
<robclark> HdkR: also, at least on some hw, halving the GPR requirements
<HdkR> That too
sagar_ has quit [Ping timeout: 480 seconds]
<Venemo> it's not clear to me why this pass works on lowered io
<bloom> d'oh
<bloom> I shouldn't be trying to remap to word-based I/O
<bloom> I should just take the location in slots and do 4*slots + component and call it a day
<bloom> and trust in the GLSL linker for linked shaders
<bloom> and then let the app run wild for SSO
<bloom> I'd still need to offset by N for the driver internal stuff
<bloom> but determining that N is a lot easier than linking all the varyings manually (and incorrectly, at this point)
Duke`` has joined #dri-devel
aleixPetitCartro has joined #dri-devel
<bloom> so.. true, the driver_location handling is totally busted
<bloom> but it's also #not_my_problem
aleixPetitCartro has quit [Remote host closed the connection]
<bloom> arguably if we just s/driver_location/handle/ a lot of confusion would go away
valde1 has joined #dri-devel
valde1 has quit [autokilled: Suspected spammer. Mail support@oftc.net with questions (2021-06-13 04:28:40)]
sdutt has joined #dri-devel
luzipher__ has joined #dri-devel
luzipher_ has quit [Ping timeout: 480 seconds]
luzipher_ has joined #dri-devel
luzipher__ has quit [Ping timeout: 480 seconds]
tarceri_ has quit [Read error: Connection reset by peer]
tarceri_ has joined #dri-devel
tarceri_ has quit [Remote host closed the connection]
tarceri has joined #dri-devel
NiksDev has joined #dri-devel
luzipher__ has joined #dri-devel
SanfordHoover-t has joined #dri-devel
SanfordHoover-t has quit [Remote host closed the connection]
luzipher_ has quit [Ping timeout: 480 seconds]
luzipher_ has joined #dri-devel
luzipher__ has quit [Ping timeout: 480 seconds]
mlankhorst has joined #dri-devel
alanc has quit [Remote host closed the connection]
alanc has joined #dri-devel
krumelmonster has joined #dri-devel
krumelmonster has quit [Remote host closed the connection]
andrey-konovalov has joined #dri-devel
Duke`` has quit [Ping timeout: 480 seconds]
Duke`` has joined #dri-devel
luzipher__ has joined #dri-devel
luzipher_ has quit [Ping timeout: 480 seconds]
krphop has joined #dri-devel
krphop has quit [Read error: Connection reset by peer]
medikoo-bot has joined #dri-devel
andrey-konovalov has quit [Ping timeout: 480 seconds]
medikoo-bot has quit [Remote host closed the connection]
luzipher_ has joined #dri-devel
sagar_ has joined #dri-devel
luzipher__ has quit [Ping timeout: 480 seconds]
gouchi has joined #dri-devel
rasterman has joined #dri-devel
sdutt has quit [Remote host closed the connection]
luzipher__ has joined #dri-devel
pekkari has joined #dri-devel
pekkari has quit []
robertfoss has joined #dri-devel
pekkari has joined #dri-devel
luzipher_ has quit [Ping timeout: 480 seconds]
Net147 has quit [Quit: Quit]
Net147 has joined #dri-devel
dv_ has quit [Quit: WeeChat 2.8]
dv_ has joined #dri-devel
dv_ has quit []
dv_ has joined #dri-devel
dv_ has quit []
dv_ has joined #dri-devel
andi_ has joined #dri-devel
andi_ has quit [Remote host closed the connection]
danvet has joined #dri-devel
Danct12 has joined #dri-devel
Daanct12 has quit [Ping timeout: 480 seconds]
pekkari has quit []
pcercuei has joined #dri-devel
ksshin[m|gr] has joined #dri-devel
ksshin[m|gr] has quit [Remote host closed the connection]
tobiasjakobi has quit [Remote host closed the connection]
jernej has quit [Quit: Free ZNC ~ Powered by LunarBNC: https://LunarBNC.net]
vimart has joined #dri-devel
vimart has quit [Remote host closed the connection]
pekkari has joined #dri-devel
<daniels> moving the very last bits of gitlab.fd.o to our new hosting is still underway and this is going to blow through our 11am UTC estimate; if you're hanging on to see it come back, please instead go for a walk in the sun, make yourself a really nice breakfast, read the 234987 tabs you've promised yourself you'll definitely deal with one day, clean the kitchen, or some other kind of Sunday activity, and we should be back for midday UTC
blue__penquin has joined #dri-devel
jernej has joined #dri-devel
blue__penquin has quit []
smpl has joined #dri-devel
mlankhorst has quit [Ping timeout: 480 seconds]
<ccr> :)
divadsn has joined #dri-devel
divadsn has quit [Remote host closed the connection]
rasterman has quit []
luzipher_ has joined #dri-devel
i-garrison has quit []
luzipher__ has quit [Ping timeout: 480 seconds]
gouchi has quit [Remote host closed the connection]
<daniels> and we're back, please shout if anything is going wrong
<kisak> and there was much rejoicing (yaaaaay)
mlankhorst has joined #dri-devel
pochu has quit [Ping timeout: 480 seconds]
korozion has joined #dri-devel
korozion has quit [Remote host closed the connection]
* ccr animates some cardboard cut medieval figures cheering
i-garrison has joined #dri-devel
jolan has quit [Quit: leaving]
soreau has quit [Remote host closed the connection]
soreau has joined #dri-devel
jolan has joined #dri-devel
Lucretia has quit [Ping timeout: 480 seconds]
hch12907 has quit [Remote host closed the connection]
hch12907 has joined #dri-devel
gouchi has joined #dri-devel
sneil_ has joined #dri-devel
Lucretia has joined #dri-devel
sneil has quit [Ping timeout: 480 seconds]
Daanct12 has joined #dri-devel
Danct12 has quit [Ping timeout: 480 seconds]
clement[m|gr]1 has joined #dri-devel
clement[m|gr]1 has quit [Remote host closed the connection]
Daaanct12 has joined #dri-devel
Daanct12 has quit [Ping timeout: 480 seconds]
tobiasjakobi has joined #dri-devel
iive has joined #dri-devel
andrey-konovalov has joined #dri-devel
<mareko> I haven't seen hw where mediump improves perf
Daanct12 has joined #dri-devel
Daaanct12 has quit [Ping timeout: 480 seconds]
<robclark> mareko: all the mobile gpu's other than adreno 2xx (which didn't have mediump)?
andrey-konovalov has quit [Ping timeout: 480 seconds]
pekkari has quit []
<anholt> rpi4 really wants mediump (twice the mul throughput, and takes half the register storage and spilling is hard to do) but the mesa compiler doesn't do it yet.
Daaanct12 has joined #dri-devel
Daanct12 has quit [Ping timeout: 480 seconds]
sylware has joined #dri-devel
tobiasjakobi has quit [Remote host closed the connection]
sdutt has joined #dri-devel
<bloom> better question is what sw uses mediump and gets a benefit from it
<bloom> gfxbench, I suppose
rasterman has joined #dri-devel
sylware has quit []
<robclark> bloom: stk does (probably mostly due to android/gles port).. and in generally it is pretty common for mobile games..
bcarvalho_ has joined #dri-devel
bcarvalho has quit [Read error: Connection reset by peer]
<robclark> daniels, anholt: I think marge-bot may need to be restarted now that gitlab.fd.o is back? She seems to be in a coma
camus has quit []
<bloom> robclark: she's taking off the weekend
<bloom> work-life balance
<Venemo> 🤣
<robclark> heheh, she might have a pile of work when she wakes up on monday :-P
TTtie-M has joined #dri-devel
TTtie-M has quit [Remote host closed the connection]
<daniels> oops!
<daniels> she's been roused from her slumber
<robclark> thx daniels !
<daniels> np!
<daniels> she just had a sleep-in then a really long lunch
<robclark> heheh
cocreature has joined #dri-devel
cocreature has quit [Remote host closed the connection]
andrey-konovalov has joined #dri-devel
Daanct12 has joined #dri-devel
bcarvalho has joined #dri-devel
bcarvalho_ has quit [Read error: Connection reset by peer]
Daaanct12 has quit [Ping timeout: 480 seconds]
thellstrom1 has quit []
troygiorshev has joined #dri-devel
troygiorshev has quit [Remote host closed the connection]
luzipher__ has joined #dri-devel
luzipher_ has quit [Ping timeout: 480 seconds]
Lucretia has quit []
rasterman has quit []
Lucretia has joined #dri-devel
mcan06[m|gr] has joined #dri-devel
mcan06[m|gr] has quit [autokilled: This host has violated network policy. Contact support@oftc.net with questions (2021-06-13 20:16:12)]
tobiasjakobi has joined #dri-devel
tobiasjakobi has quit [Remote host closed the connection]
tobiasjakobi has joined #dri-devel
bmodev[m] has joined #dri-devel
bmodev[m] has quit [Remote host closed the connection]
eac has joined #dri-devel
eac has quit [Remote host closed the connection]
<airlied> daniels: did gl just tank?
<airlied> maybe it was a glitch in the matrx
<zmike> not loading for me either
<robclark> hmm, seems to be fairly recent, I was able to start a CI run a few minutes ago, but now nothing is loading
<airlied> yeah it started just after I assigned crocus to marge-bot :-P
<airlied> obv a conspiracy to stop owners of older intel hw from seeing the power of a true gallium driver
<daniels> airlied: it did, working on it
steev has joined #dri-devel
rasterman has joined #dri-devel
<zmike> he admits to the conspiracy!
<robclark> :-P
<daniels> it's true, give me classic drivers or give me death
<dcbaker> daniels: so you have chosen... Death
<airlied> death by swrastc
<daniels> dcbaker: my checkout still has classic drivers and no-one can change that until I bring the servers back, sooooo ...
<airlied> you must render all frames for the rest of your days with swrastc
* daniels backs away from the laptop
<HdkR> Ah, so this is what happens when the Empire controls the infrastructure
<daniels> going to be minimum 10 minutes, go grab yourself a cup of tea or something
andrey-konovalov has quit [Ping timeout: 480 seconds]
* zmike settles in for a quick descriptor refactoring
<robclark> can I complain about how ARB_copy_image is the worst thing ever in the meantime?
<airlied> two days later... has anyone seen zmike?
<zmike> :P
thelounge6221 has joined #dri-devel
thelounge6221 has quit [Remote host closed the connection]
agrecascino11 has joined #dri-devel
agrecascino11 has quit [Remote host closed the connection]
GloriousEggroll has joined #dri-devel
<daniels> back
unborn has joined #dri-devel
unborn has quit [Remote host closed the connection]
<robclark> \o/
* airlied kicks and reassigns to marge, wasn't going to rescue the merge to be under the 1hr timeout
* airlied is the now marge won't do anything because the pipeline is already created
gouchi has quit [Remote host closed the connection]
<daniels> airlied: no, it's fine
<daniels> I killed the stuck jobs and restarted them, so it will complete just fine
<airlied> daniels: it was at 5x minutes runtime, so I think it was going to topple over before it ran all the armt ests
<daniels> yeah, it was, so I restarted
<airlied> granted I've now dug a bigger hole, next time I'll just let things time out
<daniels> (and cancelled to begun)
<daniels> no it's fine, just leave it and I'll shepherd it through
<airlied> cool I'm going to run away and come back in an hour and hope it's either merged or very obviously broken :-P
tobiasjakobi has quit [Remote host closed the connection]
Ultrasauce is now known as sauce
smpl has quit [Quit: smpl]
Daaanct12 has joined #dri-devel
DrNick has quit []
sdutt has quit [Remote host closed the connection]
sdutt has joined #dri-devel
Duke`` has quit [Ping timeout: 480 seconds]
Daanct12 has quit [Ping timeout: 480 seconds]
danvet has quit [Ping timeout: 480 seconds]
rasterman has quit []
luzipher_ has joined #dri-devel
luzipher__ has quit [Ping timeout: 480 seconds]
d42 has joined #dri-devel
d42 has quit [Remote host closed the connection]
iive has quit []
<bloom> robclark: complain away
<airlied> daniels: thx, the eagle has landed
<zmike> \o/
pcercuei has quit [Quit: dodo]
i-garrison has quit []
DrNick has joined #dri-devel
aswar002 has joined #dri-devel
marco44 has joined #dri-devel
marco44 has quit [Remote host closed the connection]
aswar002 has quit [Ping timeout: 480 seconds]