ChanServ changed the topic of #asahi-gpu to: Asahi Linux GPU development (no user support, NO binary reversing) | Keep things on topic | GitHub: https://alx.sh/g | Wiki: https://alx.sh/w | Logs: https://alx.sh/l/asahi-gpu
possiblemeatball1 has joined #asahi-gpu
possiblemeatball has quit [Read error: Connection reset by peer]
pthariensflame has joined #asahi-gpu
possiblemeatball1 has quit [Read error: Connection reset by peer]
possiblemeatball has joined #asahi-gpu
kidplayer666 has quit [Quit: Connection closed for inactivity]
alyssa has joined #asahi-gpu
<alyssa> i509vcb: there are lots of NaN values
<alyssa> and the hw canonicalizes
<alyssa> so NaNs don't roundtrip
<alyssa> which is why you have to cast float formats to e.g. uint on both sides of the copy
<alyssa> ditto for snorm since there are two -1.0
<alyssa> vulkan copies are supposed to be a memcpy, bit exact
<i509vcb> What's odd here is I am casting both sides of the copy
<alyssa> :S
<i509vcb> All four combinations of casts seem to do this
<alyssa> agxdecode?
<i509vcb> I'll grab that in a bit, gotta do something
<alyssa> thanks
* alyssa happy to take a look at the trace
pthariensflame has quit [Quit: Textual IRC Client: www.textualapp.com]
maximbaz has quit [Quit: bye]
<i509vcb> alyssa: ^
chadmed has quit [Quit: Konversation terminated!]
crabbedhaloablut has joined #asahi-gpu
marvin24_ has joined #asahi-gpu
marvin24 has quit [Ping timeout: 480 seconds]
john-cabaj has quit [Ping timeout: 480 seconds]
chadmed has joined #asahi-gpu
possiblemeatball has quit [Quit: Quit]
chadmed has quit [Remote host closed the connection]
kidplayer666 has joined #asahi-gpu
DarkShadow44 has quit [Quit: ZNC - https://znc.in]
DarkShadow44 has joined #asahi-gpu
chadmed has joined #asahi-gpu
cr1901 has quit [Read error: Connection reset by peer]
cr1901 has joined #asahi-gpu
alethkit has quit [Remote host closed the connection]
okt has quit [Read error: Connection reset by peer]
ella-0 has quit [Remote host closed the connection]
dorkbutt has quit [Read error: Connection reset by peer]
tsujp has quit [Remote host closed the connection]
pldtf has quit [Remote host closed the connection]
coder_kalyan has quit [Read error: Connection reset by peer]
akspecs has quit [Write error: connection closed]
kuruczgy has quit [Remote host closed the connection]
handlerug has quit [Remote host closed the connection]
d4ve has quit [Remote host closed the connection]
amada95 has quit [Remote host closed the connection]
rosefromthedead has quit [Write error: connection closed]
amada95 has joined #asahi-gpu
handlerug has joined #asahi-gpu
alethkit has joined #asahi-gpu
kuruczgy has joined #asahi-gpu
rosefromthedead has joined #asahi-gpu
coder_kalyan has joined #asahi-gpu
akspecs has joined #asahi-gpu
pldtf has joined #asahi-gpu
d4ve has joined #asahi-gpu
tsujp has joined #asahi-gpu
dorkbutt has joined #asahi-gpu
okt has joined #asahi-gpu
ella-0 has joined #asahi-gpu
crabbedhaloablut has quit [Read error: Connection reset by peer]
crabbedhaloablut has joined #asahi-gpu
<alyssa> i509vcb: You need to do the same snorm/float casting logic for CopyBufferToImage and for CopyImageToBuffer, otherwise those will canonicalize NaNs/-1.0 too
<alyssa> in e.g. agx.agxv_CmdEndRendering.0002 we're uploading floats as floats so the NaNs are getting canonicalized (and denorms flushed)
<alyssa> then in agx.agxv_CmdEndRendering.0012, we're downloading flots as floats, again canonicalizing/flushing
linuxgemini1 has joined #asahi-gpu
linuxgemini has quit [Ping timeout: 480 seconds]
derwolfe has joined #asahi-gpu
derwolfe is now known as herrwolfe
john-cabaj has joined #asahi-gpu
Misthios has quit [Quit: Misthios]
Misthios has joined #asahi-gpu
r0ni has joined #asahi-gpu
possiblemeatball has joined #asahi-gpu
maximbaz has joined #asahi-gpu
maximbaz has quit [Ping timeout: 480 seconds]
<i509vcb> alyssa: yep I forgot that and that fixed nearly every failure
<alyssa> i509vcb: awesome!
<i509vcb> Sadly a few formats always fail, I think the hardware has no SINT/UINT equivalent for these
<j`ey> i509vcb: nice going!
<i509vcb> b8g8r8a8_snorm can't be copied via meta-copy
<i509vcb> r32g32b32a32_sfloat still fails for some weird reason
<i509vcb> Although r32g32b32a32_sfloat has a hardware equvilant
skipwich has quit [Remote host closed the connection]
skipwich has joined #asahi-gpu
<i509vcb> https://pixfmtdb.emersion.fr/VK_FORMAT_B8G8R8A8_SNORM seems to show it's not much and Zink docs doesn't seem to that format is needed
AnuthaDev has joined #asahi-gpu
maximbaz has joined #asahi-gpu
<i509vcb> stupid typo for r32g32b32a32
<i509vcb> Pass: 8754, Fail: 372, Skip: 41386, Flake: 2, Duration: 3:10, Remaining: 0
<i509vcb> For dEQP-VK.api.copy_and_blit.core.image_to_image.*
<i509vcb> (3d and stencil/depth are the largest groups I need to get fixed still)
<j`ey> there's 50k tests in that subgroup?? or in total
<i509vcb> grepped out image_to_image in the api subgroup
<alyssa> j`ey: the vulkan cts is.... intense.
<j`ey> dayum
<alyssa> the full thing is millions of tests
<alyssa> ..and growing
<i509vcb> It will probably take several hours to run when agxv is done enough to try to do a full cts run
<alyssa> sigh
<j`ey> i509vcb: what machine are you using?
<i509vcb> M2 Air with 16GB of ram
<i509vcb> Ah right compressed formats, let me see how many failures I get again if I turn those back on
AnuthaDev has quit []
maximbaz has quit [Quit: bye]
possiblemeatball has quit [Quit: Quit]
<alyssa> i509vcb: what's wrong with bgra_snorm?
<alyssa> should be compatible with e.g. rgba8_unorm?
<i509vcb> I'll need to check (won't I need a swizzle for that), but I guess for that case it would make sense to have getting the copy format for snorm and sfloat formats be driver specific
thevar1able_ has quit [Remote host closed the connection]
thevar1able_ has joined #asahi-gpu
thevar1able_ has quit [Remote host closed the connection]
thevar1able_ has joined #asahi-gpu
<i509vcb> hmm no swizzle needed
crabbedhaloablut has quit []
WindowPa- has joined #asahi-gpu
WindowPain has quit [Ping timeout: 480 seconds]
possiblemeatball has joined #asahi-gpu
<i509vcb> vk_meta_draw_volume seems to do an instance per layer, I guess that's whats stopping 3D from working since I see nothing related to getting the instance number in the shader so far
handlerug has left #asahi-gpu [#asahi-gpu]
darkapex has quit [Remote host closed the connection]
darkapex has joined #asahi-gpu
<i509vcb> can't get the instance id in a fragment shader...
pthariensflame has joined #asahi-gpu
pthariensflame has quit []
<alyssa> i509vcb: vk copies are like memcpys
<alyssa> no swizzling
<alyssa> i509vcb: load the layer ID
<alyssa> vk_meta_blit_resolve uses nir_load_layer_id
<alyssa> which we'll need to lower to load_input() in the driver
<alyssa> s/the driver/agxv/
<alyssa> grep nouveau/vulkan for load_layer_id to see how
lawrence has quit [Remote host closed the connection]