09:20 rcv11x: hello
09:22 rcv11x: Is version 25.3.1 being released today? Yesterday, I played Control for the first time at 1440p with everything set to ultra and ray tracing enabled, and after playing for a few minutes, my entire system froze. Nothing worked, and I had to restart my PC. I use KDE Plasma as my desktop and an RX 9070XT. I would like to know if 25.3.1 will fix this.
09:23 K900: That's very vague, so the answer is "who knows"
09:25 pendingchaos: https://docs.mesa3d.org/release-calendar.html says 25.3.1 is being released today, but I have no idea when it will reach whatever distribution you're using
09:27 pendingchaos: it will contain "aco/insert_nops: Consider s_setpc target susceptible to VALUReadSGPRHazard", which the description says has fixed RDNA4 RT hangs with Control, so it's possible it will fix it
09:28 soreau: rcv11x: it's easy enough to check, just install mesa git or 25.3.1 to a nonstandard prefix, point LD_LIBRARY_PATH at it when starting the control game and find out
09:30 pendingchaos: VK_DRIVER_FILES is the env var to use for Vulkan if you aren't using -Dvulkan-manifest-per-architecture=false (which only exists in mesa-git)
09:31 pendingchaos: https://docs.mesa3d.org/install.html#running-against-a-local-build-easy-way
09:31 soreau: either that, or try AMD_DEBUG=usellvm, assuming it's aco-specific
09:38 glehmann: llvm doesn't support ray tracing at all
09:40 rcv11x: Ah, alright, well, I hope they fix it in version 25.3.1. For now, I'll wait for it to be released in the next few days. Thanks for the replies.
13:48 emersion: daniels: so should we Just Merge It™?
13:51 daniels: emersion: YES
13:53 emersion: who gets to push the big green button? :P
13:54 emersion: we said drm-misc-next right?
14:00 daniels: yes and I'm super happy for you to if you like?
14:04 emersion: ow Louis spotted some checkpatch warnings
15:28 mlankhorst: amd is broken in drm-tip?
15:29 mlankhorst: ddf055b80a54 ("drm/ttm: rework pipelined eviction fence handling")
15:29 mlankhorst: Most likely..
22:36 airlied: actually something else later on nukes it, but it would reduce some blocks if I got rid of it earlier
22:38 airlied: ah it gets turned into a inverse_ballot and gets constant folded
22:42 airlied: this shader is nuts, it has 20000 phis of address calculations
22:42 airlied: sorry not phis, ssa values
22:44 emersion: daniels: it's done
22:46 daniels: emersion: \o/ \o/ \o/
22:46 daniels: thankyou!
22:46 daniels: and congrats to everyone who contributed :D
22:46 emersion: yup, thanks everybody :D
22:51 airlied: poor aco gets rather upset and eats cpu tie
23:12 austriancoder: mareko: any hint why such a change in threaded_context https://gitlab.freedesktop.org/austriancoder/mesa/-/commit/321581ea024bd5fff615e49c04a12b18d1659a39 has a change in CI? https://gitlab.freedesktop.org/austriancoder/mesa/-/jobs/88711846