17:43 imirkin: cleaned out 8 years of dust on fan, applied some new thermal grease, looks like it's good as new.
17:44 karolherbst: :D I guess the CPU is now magically much cooler as well
17:44 imirkin: yeah... it used to run into the 90's when compiling. now it barely reached 60
17:44 imirkin: this wasn't a huge compile session though
17:45 imirkin: we'll see how it fares in the future
17:45 karolherbst: :( my new desktop reaches 85 clean
17:45 karolherbst: but the fan is crappy
17:45 karolherbst: and the CPU a beast
17:45 imirkin: this is a i7-920
17:45 imirkin: blast from the past ;)
17:45 imirkin: (hard to get 8 years of dust any other way though)
17:45 karolherbst: 130W...
17:45 karolherbst: mine has only 65W :D
17:45 karolherbst: I guess my fan is indeed just crappy
17:46 imirkin: its got the fan intel shipped it with
17:46 imirkin: running at 1800 rpm right now
17:46 karolherbst: mhh
17:46 imirkin: (hm, i think that might be *2... in bios it was like 1000 rpm)
17:46 karolherbst: weird
17:46 karolherbst: I have the intel one as well
17:46 karolherbst: and it usually oes up to 4000
17:47 imirkin: looks a lot like this: https://www.amazon.com/Intel-E97380-001-Cooling-Socket-1366/dp/B006BHT92G
17:47 karolherbst: yeah, mine as well
17:47 karolherbst: mine is a i7-8700
17:47 karolherbst: and usually keeps at 4GHz with 12 cores
17:47 imirkin: in between the fan and the heatsink was probably a 2mm-thick layer of dust.
17:48 karolherbst: or 4.1
17:50 imirkin: crappp... something in deqp killed glx again
17:50 imirkin: s/deqp/vk-gl-cts/
18:14 imirkin: pendingchaos: well i'll be damned... it works!
18:14 imirkin: well done!
18:15 imirkin: now i just have to figure out why VK-GL-CTS is crapping out
18:38 imirkin: well that was fast. temps hit 90 when building VK-GL-CTS
18:39 imirkin: but it used to hit the therm limits and get throttled. so still better =]
19:21 imirkin: pendingchaos: instead of PUSH_KICK, is using nouveau_fence_emit sufficient? that's what i already do in the nvc0_hw_get_query_result_resource case
20:41 pendingchaos: imirkin: that should be fine I think
20:41 pendingchaos: shall I submit a revised patch?
20:43 imirkin: no
20:43 imirkin: i've already fixed it up locally
20:44 pendingchaos: ok
20:48 imirkin: pendingchaos: pushed
20:49 imirkin: thanks =]
21:33 rtoo: How can I set the default performance level to "0a" on the kernel command line?
21:34 imirkin: nouveau.config=NvClkMode=10
21:35 rtoo: ty
22:19 imirkin: pendingchaos: another query-related issue we have is with counting compute shader invocations (part of ARB_pipeline_statistics_query)
22:20 pendingchaos: I guess I'll try to look into that once I'm done with the macro thing
22:20 imirkin: right now we just don't support it at all
22:26 imirkin: unfortunately i have no idea how to do that one
22:41 karolherbst: imirkin: ohh right, I was trying to look into that, but I couldn't mmt nvidia :(
22:42 karolherbst: imirkin: but I was thinking, do we really need to count the invocation on the GPU or could we just do that from mesa?
22:42 imirkin: indirect
22:43 karolherbst: I know that the CTS just tests against == 1, so I wasn't really able to test what I wrote, but there is also the case where the values are read out from a buffer
22:44 karolherbst: and what does "invoking a computer shader" mean? Just how often we dispatch compute shaders?
22:45 imirkin: the obvious interpretation would be the number of invocations
22:45 imirkin: i.e. local size * groups
22:45 karolherbst: okay, so basically just a ++ if _grid_launch is called
22:46 karolherbst: uhm launch_grid
22:46 imirkin: not ++
22:46 imirkin: + local size * groups
22:46 imirkin: and groups can come from an indirect param
22:46 karolherbst: okay, that was what I am not sure about
22:46 karolherbst: ahhh
22:46 karolherbst: makes sense
22:46 imirkin: well, that's the most obvious interpretation
22:47 imirkin: perhaps you could spec-lawyer your way out of it, dunno
22:47 karolherbst: yeah, as I said, CTS only tests against == 1
22:47 karolherbst: and launches 1x1x1x grids
22:48 imirkin: one could get the macro to append values to a buffer
22:48 karolherbst: imirkin: otherwise we just have a small GPU VRAM page and do an atomic add on that...
22:48 imirkin: and then have a compute shader that runs over it and evaluates it
22:48 karolherbst: yeah
22:48 imirkin: well, that'd be horribly inefficient.
22:48 imirkin: you don't want to run anything in the actual invocations
22:48 karolherbst: somebody said that the hardware doesn't have a counter for it?
22:49 imirkin: dunno
22:51 ZNTBXUpsachin`: ▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄ HAPPY APRIL FLOODS DAY BROUGHT TO YOU BY iяс.sцреяиетs.ояg сни sцреявоwl qusqlvrj: aphirst ▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄
22:51 ZNTBXUpsachin`: ▄▄▄▄▄▄▄▄▄▄ HAPPY APRIL FLOODS DAY BROUGHT TO YOU BY iяс.sцреяиетs.ояg сни sцреявоwl ykdlhyzz: Cork ▄▄▄▄▄▄▄▄▄▄▄▄▄▄
22:51 ZNTBXUpsachin`: ▄▄▄▄▄▄▄▄▄▄ HAPPY APRIL FLOODS DAY BROUGHT TO YOU BY iяс.sцреяиетs.ояg сни sцреявоwl znnrqqlyxl: Lekensteyn ▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄
22:51 ZNTBXUpsachin`: ▄▄▄▄▄▄▄▄▄▄▄▄ HAPPY APRIL FLOODS DAY BROUGHT TO YOU BY iяс.sцреяиетs.ояg сни sцреявоwl wewdliro: Kabouik_ ▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄
22:51 ZNTBXUpsachin`: ▄▄▄▄▄▄▄▄▄▄▄ HAPPY APRIL FLOODS DAY BROUGHT TO YOU BY iяс.sцреяиетs.ояg сни sцреявоwl usyra: NolanSyKinsley ▄▄▄▄▄▄▄▄▄▄▄▄
22:51 ZNTBXUpsachin`: ▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄ HAPPY APRIL FLOODS DAY BROUGHT TO YOU BY iяс.sцреяиетs.ояg сни sцреявоwl ctaggrute: Arbition ▄▄▄▄▄▄▄▄▄▄▄
22:51 ZNTBXUpsachin`: ▄▄▄▄▄▄▄▄▄▄ HAPPY APRIL FLOODS DAY BROUGHT TO YOU BY iяс.sцреяиетs.ояg сни sцреявоwl sftfjve: cwabbott ▄▄▄▄▄▄▄▄▄▄▄▄
22:51 ZNTBXUpsachin`: ▄▄▄▄▄▄▄▄▄▄▄▄▄ HAPPY APRIL FLOODS DAY BROUGHT TO YOU BY iяс.sцреяиетs.ояg сни sцреявоwl annfskl: halfline ▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄
22:51 ZNTBXUpsachin`: ▄▄▄▄▄▄▄▄▄▄▄▄▄▄ HAPPY APRIL FLOODS DAY BROUGHT TO YOU BY iяс.sцреяиетs.ояg сни sцреявоwl qgfinpzy: karolherbst ▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄
22:51 ZNTBXUpsachin`: ▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄ HAPPY APRIL FLOODS DAY BROUGHT TO YOU BY iяс.sцреяиетs.ояg сни sцреявоwl qjvugj: glennk ▄▄▄▄▄▄▄▄▄▄▄▄
22:51 ZNTBXUpsachin`: ▄▄▄▄▄▄▄▄▄▄▄ HAPPY APRIL FLOODS DAY BROUGHT TO YOU BY iяс.sцреяиетs.ояg сни sцреявоwl yaqylktbd: infinity0 ▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄
22:51 ZNTBXUpsachin`: ▄▄▄▄▄▄▄▄▄▄▄▄▄ HAPPY APRIL FLOODS DAY BROUGHT TO YOU BY iяс.sцреяиетs.ояg сни sцреявоwl gaieccpqr: mithro ▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄
22:51 ZNTBXUpsachin`: ▄▄▄▄▄▄▄▄▄▄ HAPPY APRIL FLOODS DAY BROUGHT TO YOU BY iяс.sцреяиетs.ояg сни sцреявоwl zonulsgn: RushPL ▄▄▄▄▄▄▄▄▄▄▄▄▄▄
22:51 ZNTBXUpsachin`: ▄▄▄▄▄▄▄▄▄▄▄▄▄▄ HAPPY APRIL FLOODS DAY BROUGHT TO YOU BY iяс.sцреяиетs.ояg сни sцреявоwl xlkrhalot: nwm ▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄
22:51 aphirst: bit sodding late fam
22:51 ZNTBXUpsachin`: ▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄ HAPPY APRIL FLOODS DAY BROUGHT TO YOU BY iяс.sцреяиетs.ояg сни sцреявоwl wmknvjhdm: Jon ▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄
22:56 imirkin: mupuf: what do i need to do to fix this (on nv34):
22:56 imirkin: ERROR: Can't get value of subfeature temp1_max: Can't read
22:56 imirkin: (fix = make the damn thing not show up at all)