04:14mareko: zmike: can you please stop using nir_io_dont_optimize?
08:22Emantor: narmstrong: Hi, I have a question regarding panel-simple: would reset gpio support be in scope for panel simple or should I break this out into a separate panel driver?
08:38mmind00: mripard: tzimmermann: question about two series living partially outside my usual drm/rockchip playground:
08:38mmind00: - https://patchwork.kernel.org/project/dri-devel/cover/20250220234141.2788785-1-heiko@sntech.de/ ... has a drivers/base element, Acked by gregkh
08:38mmind00: - https://patchwork.kernel.org/project/dri-devel/cover/20250217215641.372723-1-detlev.casanova@collabora.com/ ... has a drm/bridge element, reviewed by robertfoss (as listed bridge maintainer)
08:38mmind00: so in my reading of the drmmisc doc, I should be "allowed" to apply them to drmmisc, but want to make sure before causing unnecessary work for other people
08:39tzimmermann: mmind00, i'm fine with merging this through drm-misc. as long as it's ok for the maiantainers of drivers/base
08:40tzimmermann: you can also merge this through the other tree, if they disagree
08:41mmind00: tzimmermann: with gregkh being the maintainer of drivers/base, I assume their "Acked-by" was meant as merge-whereever, but I guess I can also ask to confirm ... for safety :-)
08:42tzimmermann: yeah, maybe ask him first
08:47narmstrong: Emantor: panel-simple already supports an enable-gpio
08:47Emantor: narmstrong: I know, but unfortunately thats not enough for this panel to work reliably
08:48narmstrong: Emantor: which gpio would you need ? reset ?
08:49Emantor: Yes, reset and than correct timing between reset and enable AFAICS.
09:09narmstrong: Emantor: Try to add it to panel-simple and send an RFC, perhaps it would fit
09:10Emantor: narmstrong: sure, thanks for the info :)
10:17fartenderman: i did some testing and found that three data access methodical ways worked, but for the last there is another way, essentially the problem with +- gate is negative and positive polarities can not be used at the same time to filter results. So there is actually a way in hashed computation world to make two banks and fixed intrinsics pinned, but today there was confusion over the article i
10:17fartenderman: posted, i comment on that instead. You see the author was so clever not to leave any open questions about this paper. He would end up naming trace entropy binary function. and it already says what to expect even if you do not know those math terms, but know cause there is only 32 cascade bus traces, so it was clever article, so you have 32 numbers from one distribution and add uniform
10:17fartenderman: set to the distribution of 32 elements in case of 32bit. cause that would match my theory, but myself i read the article one more time in case i was wrong.
10:20mareko: zmike: https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/33794
13:21jani: tzimmermann: mripard: mlankhorst: ack for merging https://lore.kernel.org/r/20250220104144.207526-2-vinod.govindapillai@intel.com via drm-intel, please?
13:22tzimmermann: jani, ack
13:22jani: tzimmermann: thanks!
13:23tzimmermann: IMHO we should be more careful about using const where possible
13:23tzimmermann: so +1 for this patch
13:52mmind00: tzimmermann: to finalize from this morning, gregkh said "Yup, merge away!" :-)
13:52tzimmermann: then go ahead
14:01zmike: mareko: working on it, but I've got a lot of stuff in motion and finding time to debug/fix compiler stuff is not super high on my priority list
15:04karolherbst: I got e65c1473ded8e278b2253b7939af7a2fbfc72714 breaking stuff :'(
15:04karolherbst: will debug later
15:10jannau: dakr: https://rust-for-linux.zulipchat.com/#narrow/channel/288089-General/topic/build.20checked.20IoMem.20access
16:36karolherbst: ahh... no...
16:36karolherbst: nir_opt_dead_write_vars is broken before things got inlined...
16:44karolherbst: I should get the CL CTS into CI 🙃
17:25alyssa: karolherbst: can you skip conversions? *sweat*
17:25karolherbst: with -w it's not _that_ bad 🙃
17:25karolherbst: images are worse
17:26karolherbst: but yeah.. I don't think I'd like to run expensive stuff
18:45milek7: is there way to force fp16 support in radv on polaris?
18:48glehmann: milek7: I think it should work if you use it (ofc we never test it but compiler support is implement)
18:50glehmann: this is the line that you would need to change to expose it: https://gitlab.freedesktop.org/mesa/mesa/-/blob/main/src/amd/vulkan/radv_physical_device.c#L867
18:50milek7: thanks, I will try that
18:51milek7: I'm trying to launch indiana jones game on rx580
19:15milek7: yay, it works
19:15milek7: with emulate_rt and fp16 patch
19:19glehmann: milek7: incredible, is it actually playable on lower settings?
19:25pac85: I'd expect it to be
19:25pac85: It does barely any rt
19:25milek7: around 30fps on low at 1080p
19:26alyssa: wow!
19:27pac85: Would be cool to see an rgp capture
19:30pac85: alyssa: indy on agx when 🙃
19:33pixelcluster: weeee
19:33pixelcluster: nice
19:40alyssa: pac85: please no
20:01milek7: pac85: MESA_VK_TRACE=rgp or something else? it writes 500MB file but RadeonGPUProfiler doesn't open it
20:02pac85: milek7: yeah I was curious how much time was spent tracing rays but if it can't open it rip
20:17milek7: http://s2.milek7.pl/TheGreatCircle.exe_2025.02.27_20.51.56.rgp
20:17milek7: it opens on rgp 1.16
20:28milek7: 20% of time is used by raytracing, if I'm looking correctly
20:46pac85: so about 9ms tracing rays and 2.2 building accell structures
20:46pac85: not bad at all