ChanServ changed the topic of #zink to: official development channel for the mesa3d zink driver || https://docs.mesa3d.org/drivers/zink.html
airlied has quit [Remote host closed the connection]
airlied has joined #zink
sinanmoh- has joined #zink
sinanmohd has quit [Ping timeout: 480 seconds]
Sachiel has quit [Quit: WeeChat 3.7.1]
Sachiel has joined #zink
airlied has quit [Ping timeout: 480 seconds]
airlied has joined #zink
airlied has quit [Ping timeout: 480 seconds]
alyssa has joined #zink
<alyssa> zmike: Does dEQP-GLES31.functional.shaders.builtin_functions.pack_unpack.unpackhalf2x16_geometry pass for you under lazapipe?
<alyssa> Getting a terrifying LLVM error but thinking it might be arm64 only
<alyssa> LLVM ERROR: Cannot select: 0xffff70034f30: v4f32 = fp_extend 0xffff70034ec0
<alyssa> yeah going to guess this is a bug in the llvm arm64 emitter
<alyssa> terrifying
<zmike> alyssa: you can see the full fail list for drivers in zink/ci
<alyssa> ack
<alyssa> so not an xfail on x86
<alyssa> https://rosenzweig.io/dubious.txt <-- the LLVM IR looks incredibly dubious in fairness
<zmike> I'll take your word for it
<alyssa> %37 = fpext <4 x i16> %36 to <4 x float>
<alyssa> this bit
<zmike> lavapipe hasn't been well tested on arm64
<zmike> to my knowledge
<alyssa> should be f16, if llvm has such a type
<zmike> I imagine it does
alyssa has left #zink [#zink]
dubiousness has quit [Remote host closed the connection]
sumoon has quit [Remote host closed the connection]
alyssa has joined #zink
sumoon has joined #zink
dubiousness has joined #zink
<alyssa> zmike: v3dz doesn't have the same ring to it
<alyssa> z3dz maybe?
<zmike> 😬
<alyssa> I went with v3dz for my shell alias
<dj-death> is anholt_ on holiday?
<zmike> seems possible
<dj-death> trying to chase a bunch of crashes with zink : https://gitlab.freedesktop.org/mesa/mesa/-/jobs/42296468
<dj-death> unfortunately only on CI
<dj-death> validation complains locally but that passes
<dj-death> so I'm wondering if the CI has an older VVL installed and maybe that crashes
<dj-death> because it gets errors too
<zmike> CI has the latest tag I think
<zmike> but also I don't think CI uses VVL for anv
alyssa has left #zink [#zink]
<dj-death> it shows in the logs
<zmike> huh
<dj-death> quite a lot of "The Vulkan spec states: If the pipeline is being created with pre-rasterization shader state and pStages includes tessellation shader stages, the topology member of pInputAssembly must be VK_PRIMITIVE_TOPOLOGY_PATCH_LIST"
<zmike> guess I forgot
<dj-death> seems like every test crashes has the same kind of error
<dj-death> "The Vulkan spec states: If the pipeline is being created with pre-rasterization shader state and vertex input state and the topology member of pInputAssembly is VK_PRIMITIVE_TOPOLOGY_PATCH_LIST, pStages must include tessellation shader stages"
<dj-death> like this one : KHR-GL46.constant_expressions.array_normalize_vec2_fragment
<zmike> 🤔
<zmike> well the VVL tag is in the gitlab scripts
<dj-death> ah lol
<dj-death> apparently that error only shows if you run a tess test before
<zmike> hm
<dj-death> try those 2:
<dj-death> KHR-GL46.constant_expressions.array_dot_float_tess_control
<dj-death> KHR-GL46.constant_expressions.array_normalize_vec2_fragment
<zmike> when I get home I can
<dj-death> thanks
<zmike> hm I guess I need your MR to see this?
airlied has joined #zink
<zmike> huh
<zmike> I didn't know anv supported dynamicPrimitiveTopologyUnrestricted
<zmike> anyway this looks like a combo spec bug and vvl problem
<zmike> dj-death: I'd recommend just adding a vvl exception to ci
<zmike> I'll file spec issues
<zmike> confusing why this doesn't trigger on lavapipe though
<dj-death> zmike: so the CI aborts or something on validation errors?
<zmike> yes
<dj-death> okay, I didn't know that
<zmike> ah ok
i509vcb has joined #zink
jabashque_ has quit []