<alyssa>
Venemo_: pendingchaos: OOI have you looked into the above? (NIR+ACO performance with abs/neg/saturate and nir_register/write_mask/is_ssa checks/etc deleted?
<alyssa>
)
<pendingchaos>
I have not
<alyssa>
oh, also the massive ALU swizzles
<alyssa>
(which are only ever used for OpenCL and get lowered before reaching the backend on every backend)
<alyssa>
pendingchaos: you did the gc_sweep right? might be an interesting data point to get
FloGrauper[m] has joined #dri-devel
co1umbarius has joined #dri-devel
heat has quit [Remote host closed the connection]
heat has joined #dri-devel
columbarius has quit [Ping timeout: 480 seconds]
yuq825 has joined #dri-devel
macromorgan is now known as Guest1182
Guest1182 has quit [Read error: Connection reset by peer]
<jenatali>
Nope. You're probably not authed with NickServ
Florian1 has joined #dri-devel
YuGiOhJCJ has quit [Remote host closed the connection]
YuGiOhJCJ has joined #dri-devel
ybogdano has quit [Ping timeout: 480 seconds]
Leopold has quit []
FloGrauper[m] has quit []
FloGrauper[m] has joined #dri-devel
Leopold_ has joined #dri-devel
<pinchartl>
and another i915 GPU freeze requiring a hard reboot :-(
<pinchartl>
that's an average of 2-3 per month
<jenatali>
Flo Grauper: No, auto-auth doesn't work on the OFTC bridge, you need to manually message NickServ
lemonzest has joined #dri-devel
<jenatali>
I don't remember how I initiated a chat unfortunately, that was a couple years ago that I've just had a chat open for re-auth lol
bcheng has quit [Remote host closed the connection]
bcheng has joined #dri-devel
<FloGrauper[m]>
this is to check if registration worked
<FloGrauper[m]>
hi there. im slightly confused about whether anv is supposed to advertise VK_KHR_ray_tracing_pipeline on DG2 gpus. afaict it should be supported as of 647773a0 but it doesnt show up in vulkaninfo for me. should it?
<HdkR>
FloGrauper[m]: Intel RT feature needs intel-clc to be enabled, which arch mesa-git doesn't enable. It's disabled by default.
Leopold_ has quit [Remote host closed the connection]
Daanct12 has joined #dri-devel
Daanct12 has quit [Remote host closed the connection]
maxzor has quit [Ping timeout: 480 seconds]
OftenTimeConsuming has quit [Remote host closed the connection]
OftenTimeConsuming has joined #dri-devel
anholt_ has quit [Remote host closed the connection]
anholt has joined #dri-devel
<Lynne>
airlied: ...300 lines of pure mindless tourment later, I get it to not crash
<Lynne>
seriously, there's overdoing it, there's throwing the kitchen sink, and then there's the pure wall of values from headers you _have_ to set to decode HEVC
<FloGrauper[m]>
<HdkR> "Flo Grauper: Intel RT feature..." <- thanks. got it to work after some arguing with meson
<Lynne>
we used to joke in the aom wg that av1's uncompressed header was massive, but this is on a whole different level
ngcortes has quit [Read error: Connection reset by peer]
<FloGrauper[m]>
<HdkR> "Flo Grauper: Intel RT feature..." <- figured it quickly try quake rtx, getting `FATAL: Could not find a suitable Vulkan queue family!` i assume thats just broken at the moment?
<FloGrauper[m]>
if so, is there any visual demo that works?
<HdkR>
I don't know if it is in a suitable state for use
<FloGrauper[m]>
i get that impression
<FloGrauper[m]>
(that its not)
tarceri has quit [Remote host closed the connection]
tarceri has joined #dri-devel
tarceri has quit [Remote host closed the connection]
tarceri has joined #dri-devel
aravind has joined #dri-devel
tarceri has quit [Remote host closed the connection]
tarceri has joined #dri-devel
tarceri has quit [Remote host closed the connection]
tarceri has joined #dri-devel
tarceri has quit [Remote host closed the connection]
tarceri has joined #dri-devel
heat has quit [Ping timeout: 480 seconds]
tarceri has quit [Remote host closed the connection]
tarceri has joined #dri-devel
tarceri has quit [Remote host closed the connection]
<kurufu>
Seems there are fixes in the github so you could try building it manually and running it.
tarceri has quit [Remote host closed the connection]
tarceri has joined #dri-devel
tarceri has quit [Remote host closed the connection]
tarceri has joined #dri-devel
tarceri has quit [Remote host closed the connection]
tarceri has joined #dri-devel
tarceri has quit [Remote host closed the connection]
tarceri has joined #dri-devel
slattann has joined #dri-devel
tarceri has quit [Remote host closed the connection]
tarceri has joined #dri-devel
tarceri has quit [Remote host closed the connection]
tarceri has joined #dri-devel
<kurufu>
actually i just built it myself and their fix is still broken for single queue support. However they also depend on sparsebinding which doesnt seem to be supported in anv.
<FloGrauper[m]>
<kurufu> "Flo Grauper: https://gitlab...." <- Radv is working fine for me though
<kurufu>
odd, doesnt work for me on arc. Though it also seems like the sparseBinding request is actually a mistake and sparseBinding is not used.
<kurufu>
anyway hacking on the github and allowing the graphics queue to be the transfer queue and removing the request for sparseBinding it runs and maybe renders correctly. Though its chugging along at 25fps on my 750.
<FloGrauper[m]>
Arc with anv doesn’t work for me either. Navi 23 works with both radv and amdvlk
<kurufu>
Thanks for asking, i was also wondering about playing with a raytracing demo but didnt know of any hah.
jewins has joined #dri-devel
Jeremy_Rand_Talos_ has joined #dri-devel
<Lynne>
airlied: pushed my pretty-complete hevc code
<Lynne>
still doesn't work, displays corruption even with I-frames
<Lynne>
sets all parameters it should need, sps, pps and vps
<Lynne>
though I may have missed some, there are far too much flags
kts has joined #dri-devel
<Lynne>
also there's an issue with vkvps_nal_hdr, both variables should be [HEVC_MAX_VPS_COUNT][HEVC_MAX_LAYER_SETS][HEVC_SUBLAYERS_LIST_SIZE] but chaning it for some reason segfaults
<Lynne>
I'll go pass out and hope I don't have nightmares about mindlessly setting and checking hundreds of fields and flags
<airlied>
Lynne: cool, ive been out sick today, will see if tomorrow gives me a clear enough head to debug it further
<airlied>
i only tested an i frame with the cts
danvet has joined #dri-devel
fab has joined #dri-devel
jewins has quit [Ping timeout: 480 seconds]
kts has quit [Quit: Leaving]
fab has quit [Quit: fab]
kts has joined #dri-devel
bgs has quit [Remote host closed the connection]
dcz_ has joined #dri-devel
rmckeever has quit [Quit: Leaving]
alanc has quit [Remote host closed the connection]
alanc has joined #dri-devel
slattann has quit [Ping timeout: 480 seconds]
macromorgan has quit [Read error: Connection reset by peer]
tursulin has joined #dri-devel
djbw has quit [Read error: Connection reset by peer]
fab has joined #dri-devel
Duke`` has joined #dri-devel
mvlad has joined #dri-devel
slattann has joined #dri-devel
rasterman has joined #dri-devel
frieder has joined #dri-devel
djbw has joined #dri-devel
MajorBiscuit has joined #dri-devel
sgruszka has joined #dri-devel
Major_Biscuit has joined #dri-devel
rgallaispou has quit [Read error: Connection reset by peer]
MajorBiscuit has quit [Ping timeout: 480 seconds]
camus has joined #dri-devel
lynxeye has joined #dri-devel
sarahwalker has joined #dri-devel
shashanks has joined #dri-devel
DottorLeo has joined #dri-devel
<DottorLeo>
hi!
<Nyaa>
hello
<DottorLeo>
Technically, what maximum version of Vulkan can be supported on Southern Island/Volcan Island?
Company has joined #dri-devel
pcercuei has joined #dri-devel
kts has quit [Remote host closed the connection]
maxzor has joined #dri-devel
Duke`` has quit [Ping timeout: 480 seconds]
<Nyaa>
DottorLeo, I do not see any reason they can't support Vulkan 1.3 Core, no idea about extensions though
fab has quit [Read error: Connection reset by peer]
fab_ has joined #dri-devel
fab_ is now known as Guest1222
<Nyaa>
Vulkan 1.3 Core can run on any GPU capable of running OpenGL ES 3.1, so mostly comes down to if someone will bother implementing it or not
maxzor has quit [Ping timeout: 480 seconds]
<DottorLeo>
Nyaa, thanks
Jeremy_Rand_Talos__ has joined #dri-devel
<Nyaa>
GCN 1.0 and above all support ES 3.1 to my knowledge
Jeremy_Rand_Talos_ has quit [Remote host closed the connection]
<Nyaa>
possibly even TeraScale 2
devilhorns has joined #dri-devel
DottorLeo has quit [Quit: Leaving]
heat has joined #dri-devel
OftenTimeConsuming has quit [Remote host closed the connection]
rgallaispou has joined #dri-devel
OftenTimeConsuming has joined #dri-devel
<mriesch>
can panels that are partially occluded be handled in drm?
<mriesch>
maybe by reporting the effective resolution or similar?
<ishitatsuyuki>
is this a M2 Mac reference ;)
Jeremy_Rand_Talos_ has joined #dri-devel
<mriesch>
M2 Mac?
<jannau>
mriesch: m1 macbook pro and m2 macbook air have a "notch" with a camera inside the main display
Jeremy_Rand_Talos__ has quit [Ping timeout: 480 seconds]
<mriesch>
ah, got it :-) well no, not using mac at all (apart from that thing that connects my soc to ethernet)
<jannau>
asahi linux is currently dealing with it by ignoring the partially occluded top edge.
<mriesch>
what i am dealing with is a panel whose visible area is restricted due to reasons of mounting/construction
<jannau>
the top 64/74 pixels are ignored. this is currently hardcoded but I think nothing would prevent userspace from doing the same
maxzor has joined #dri-devel
<mriesch>
jannau: hardcoded where exactly?
<mriesch>
i was thinking that mostly this is hardware related and should be modeled in e.g. device tree or so
<jannau>
in the bootloader for the boot framebuffer and in the out-of-tree drm/kms driver
<mriesch>
jannau: ah ok...
<jannau>
not really hardcoded in the driver, the bootloader adds the vendor property "apple,notch-height" and the driver uses that ignore the specified amount of pixels at the top edge
<mriesch>
and by ignored you mean that pixels are drawn in this particular region but they are simply black or something?
<mriesch>
or is the panel size effectively reduced?
<emersion>
the goal is to figure out the best way to describe notches
<emersion>
then expose that in KMS, then later in Wayland
<mriesch>
emersion: thanks, i knew i heard this notch thing before somewhere but couldn't recall in which context
<emersion>
stalled because lack of manpower mainly
<jannau>
the reported mode is smaller than the panel size and scanout does not begin at the top of the panel. the display processor interface is allows that in our case
<emersion>
that's an ok-ish workaround, but a proper solution would be better, jannau
<mriesch>
notches are more complicated i guess, for my case such an approach could work
<mriesch>
instead of having the full 320x240 i have (effectively) an 280x240 display, but it is still a rectangle (no notches, no curved edges)
<jannau>
emersion: agreed, there are unfortunately more than enough issues in the display driver look more important than proper notch handling
<jannau>
we should look at it though since there will be long delay until it is used in userspace
Duke`` has joined #dri-devel
Guest1222 has quit [Read error: Connection reset by peer]
fab has joined #dri-devel
warpme____ has joined #dri-devel
sgruszka has quit [Ping timeout: 480 seconds]
Akari has joined #dri-devel
alarumbe has quit [Quit: ZNC 1.8.2+deb2 - https://znc.in]
kts has joined #dri-devel
<qyliss>
Is it likely to be intentional that -Dgallium-drivers=auto,zink or -Dgallium-drivers=auto,asahi won't actually enable zink or asahi?
<qyliss>
(Asking in case there's some reason it wouldn't make sense to use either of those in addition to the auto set)
bgs has joined #dri-devel
sgruszka has joined #dri-devel
tango_ has quit [Ping timeout: 480 seconds]
jkrzyszt has joined #dri-devel
srslypascal is now known as Guest1230
srslypascal has joined #dri-devel
kts_ has joined #dri-devel
kts has quit [Ping timeout: 480 seconds]
<Lynne>
airlied: I found one issue, the reason why it crashed
Guest1230 has quit [Ping timeout: 480 seconds]
<Lynne>
the number of bytes needed to describe all flags and pps/sps/vps parameters is 119511040
<Lynne>
that's 113 megabytes!
<Lynne>
113 megabytes of setting fields just to decode a single frame!
agd5f has quit [Read error: Connection reset by peer]
<Lynne>
sure, the video wg would like to leave all parsing of data up to the user, but still, 113 megabytes of data getting copied field by field at least 3, maybe 4 times
<Lynne>
most of it is eaten by the vps header data, since you have 16 max vps, each of which can have 1024 layers, and each layer can have 7 sublayers
<Lynne>
each header of which is 132 bytes
sgruszka has quit [Ping timeout: 480 seconds]
tango_ has joined #dri-devel
devilhorns has quit []
noord has quit [Quit: bye]
Akari has quit [Quit: segmentation fault (core dumped)]
noord has joined #dri-devel
YuGiOhJCJ has quit [Quit: YuGiOhJCJ]
maxzor has quit [Ping timeout: 480 seconds]
kts_ has quit []
tzimmermann has joined #dri-devel
tzimmermann has quit [Remote host closed the connection]
tzimmermann has joined #dri-devel
Leopold_ has joined #dri-devel
maxzor has joined #dri-devel
agd5f has joined #dri-devel
<ajax>
qyliss: i doubt it's intentional. i'd call it broken, even.
<qyliss>
cool, I'll make an MR then :)
jkrzyszt has quit [Remote host closed the connection]
yuq825 has quit []
flto has quit [Remote host closed the connection]
kts has joined #dri-devel
flto has joined #dri-devel
jewins has joined #dri-devel
jkrzyszt has joined #dri-devel
kts has quit [Quit: Leaving]
kts has joined #dri-devel
<Lynne>
airlied: HEVC code works on nvidia!
bcheng has quit [Remote host closed the connection]
bcheng has joined #dri-devel
degasus has quit [Ping timeout: 480 seconds]
Haaninjo has joined #dri-devel
aravind has quit [Ping timeout: 480 seconds]
fab has quit [Quit: fab]
fxkamd has joined #dri-devel
Haaninjo has quit [Read error: Connection reset by peer]
Haaninjo has joined #dri-devel
gawin has joined #dri-devel
pjakobsson has joined #dri-devel
off^ has quit [Remote host closed the connection]
dliviu has quit [Ping timeout: 480 seconds]
fab has joined #dri-devel
camus has quit [Ping timeout: 480 seconds]
Major_Biscuit has quit []
jkrzyszt has quit [Remote host closed the connection]
hwentlan_ has left #dri-devel [#dri-devel]
mbrost has joined #dri-devel
pekkari has joined #dri-devel
i-garrison has quit [Ping timeout: 480 seconds]
jkrzyszt has joined #dri-devel
bgs has quit [Remote host closed the connection]
sarahwalker has quit [Remote host closed the connection]
i-garrison has joined #dri-devel
jkrzyszt has quit [Remote host closed the connection]
kts has quit [Quit: Leaving]
mbrost has quit [Remote host closed the connection]
mbrost has joined #dri-devel
jkrzyszt has joined #dri-devel
ybogdano has joined #dri-devel
lynxeye has quit [Quit: Leaving.]
pekkari has quit [Quit: Konversation terminated!]
mbrost has quit [Ping timeout: 480 seconds]
mbrost has joined #dri-devel
gawin has quit [Ping timeout: 480 seconds]
jeeeun8413 has quit []
jeeeun8413 has joined #dri-devel
tursulin has quit [Ping timeout: 480 seconds]
tzimmermann has quit [Quit: Leaving]
jkrzyszt has quit [Remote host closed the connection]
bbrezillon has quit [Ping timeout: 480 seconds]
dliviu has joined #dri-devel
gawin has joined #dri-devel
frieder has quit [Remote host closed the connection]
junaid has joined #dri-devel
jkrzyszt has joined #dri-devel
Akari has joined #dri-devel
bbrezillon has joined #dri-devel
Nyaa has quit [Ping timeout: 480 seconds]
Akari` has joined #dri-devel
Akari has quit [Remote host closed the connection]
alarumbe has joined #dri-devel
jkrzyszt has quit [Remote host closed the connection]
jkrzyszt has joined #dri-devel
mbrost has quit [Ping timeout: 480 seconds]
gouchi has joined #dri-devel
shashanks has quit [Ping timeout: 480 seconds]
shashanks has joined #dri-devel
macromorgan has joined #dri-devel
phire has quit [Remote host closed the connection]
ngcortes has joined #dri-devel
phire has joined #dri-devel
eukara has joined #dri-devel
pjakobsson_ has joined #dri-devel
<dwlsalmeida>
hi, I am trying to hack together userptr support for the va frontend mainly for learning purposes. The va frontend uses resource_from_handle with WINSYS_HANDLE_TYPE_FD to import a dmabuf. For userptr though I am using resource_from_user_memory. Problem is that vcn apparently expects a si_texture to be created, which happens as part of
<dwlsalmeida>
resource_from_handle (which is set to si_texture_from_handle), but doesn't with resource_from_user_memory (which is set to si_buffer_from_user_memory), am I missing something?
pjakobsson has quit [Ping timeout: 480 seconds]
Akari` has quit []
<dwlsalmeida>
the default VA_MEMORY path will also create a si_surface through vl_video_buffer_create_as_resource, by calling resource_create_with_modifiers, which is set to si_texture_create_with_modifiers
<dwlsalmeida>
si_texture, i mean
<agd5f>
dwlsalmeida, you can't export userptr memory via dma-buf IIRC
<agd5f>
that might not be relevant depending on what you are doing
<dwlsalmeida>
I want the gpu to decode directly into userspace memory, which on intel works through DRM_IOCTL_I915_GEM_USERPTR. I managed to create a pipe_resource from a user pointer by using resource_from_user_memory(), which internally calls amdgpu_create_bo_from_user_mem(), but now my question is why can't I create a struct si_texture out of that
<dwlsalmeida>
pipe_resouce in radeonsi (apparently?)
<dwlsalmeida>
is it just a matter of adding more code, or is this not supposed to work in the first place?
slattann has quit [Ping timeout: 480 seconds]
<dwlsalmeida>
a second question would be whether it's possible to force the gpu to decode to a userspace buffer in a linear layout instead of tiled, otherwise I'd have to detile it somehow
kem has quit [Ping timeout: 480 seconds]
rasterman has quit [Quit: Gettin' stinky!]
<FloGrauper[m]>
<kurufu> "https://gist.github.com/..."; <- as an update on that: i applied that to https://github.com/NVIDIA/Q2RTX/pull/221 and im now getting `validation layer 1 4096: execbuf2 failed: Cannot allocate memory (VK_ERROR_DEVICE_LOST)` im currently trying to figure out how to make the master not try to build the broken wayland support
<kurufu>
i just disabled the wayland support in cmake for sdl
<kurufu>
i was on the demo pak if that matters, its just what the steam version had auto selected when you installed it.
<FloGrauper[m]>
i dont think the error im getting is related to wayland
mvlad has quit [Remote host closed the connection]
kem has joined #dri-devel
alarumbe has quit [Quit: ZNC 1.8.2+deb2 - https://znc.in]
<airlied>
Lynne: seeing a crash in vk_hevc_fill_pict/ff_vk_decode_prepare_frame
<airlied>
pic->hwaccel_picture_private seems NULL
<Lynne>
use intra-only for now
<Lynne>
I'm almost done with the missing chunk needed to handle refs
Lucretia has quit [Remote host closed the connection]
Lucretia has joined #dri-devel
rmckeever has joined #dri-devel
ybogdano has quit [Ping timeout: 480 seconds]
alyssa has left #dri-devel [#dri-devel]
mbrost has joined #dri-devel
<airlied>
Lynne: ah I'll find an I video :-P
<Lynne>
-c:v libx265 -keyint_min 1 -g 1
<Lynne>
it'll even automatically set the profile to hevc intra
<Lynne>
just make sure to either use 8bit stuff or -pixfmt yuv420p
rsripada has quit []
mbrost has quit [Ping timeout: 480 seconds]
mbrost has joined #dri-devel
wizard5623 has joined #dri-devel
jkrzyszt has quit [Remote host closed the connection]
rsripada has joined #dri-devel
wizard5623 has left #dri-devel [Leaving]
fab has quit [Quit: fab]
junaid has quit [Remote host closed the connection]
mbrost has quit [Ping timeout: 480 seconds]
Duke`` has quit [Ping timeout: 480 seconds]
mbrost has joined #dri-devel
dcz_ has quit [Ping timeout: 480 seconds]
danvet has quit [Ping timeout: 480 seconds]
Nyaa has joined #dri-devel
<Nyaa>
well, I may not have tracked down broken 3d in r300 big endian, but i did manage to track down an assertion error and possibly why glxgears corrupts on newer versions when notcl is set
Nyaa has quit [Remote host closed the connection]
Nyaa has joined #dri-devel
<Nyaa>
[sending again because I lost connection lol] well, I may not have tracked down broken 3d in r300 big endian, but i did manage to track down an assertion error and possibly why glxgears corrupts on newer versions when notcl is set
<Nyaa>
now i just gotta compile one last time to double check
mbrost has quit [Ping timeout: 480 seconds]
ybogdano has joined #dri-devel
gawin has quit [Ping timeout: 480 seconds]
agd5f has quit [Remote host closed the connection]
agd5f has joined #dri-devel
<Lynne>
airlied: any progress getting intra to work? I did get a few frames last night
<Lynne>
they were partially corrupt, but half the frame was good
<Lynne>
also, I appear to be getting visible free'd GPU memory at times
<airlied>
still getting wierd crash, I expcet my command line isn't good
pcercuei has quit [Read error: Connection reset by peer]
pcercuei has joined #dri-devel
<DemiMarie>
alyssa: what about those (e.g. in scientific computing) that need bit-for-bit reproducibility?
<DemiMarie>
(A lot of physical systems are chaotic, so even tiny differences can be exponentially amplified. Therefore, if one wants to be able to reproduce results, they need to be bit-for-bit reproducible.)
<Lynne>
airlied: yeah, I think it's probably either a problem with the parsing or a bitstream desync somehow