02:20karolherbst: imirkin: according to glisse the push buffer engines only has support for 40 bit pointers... no quite sure what he means by that, but I guess it's the address of the push buffers
04:55pmoreau: karolherbst: I’ll see what I can do for the MR.
04:56pmoreau: Also, the OpenCL spec will finally say something regarding BuiltIn variables' StorageClass: https://github.com/KhronosGroup/OpenCL-Docs/pull/278
10:33karolherbst: pmoreau: is this already fixed in the translator or do we need to fix this?
10:51pmoreau: It is already fixed in the translator, for about a year already I think.
10:52pmoreau: You were even the one who sent in the MR for it, IIRC. ;-)
10:58karolherbst: okay :D
10:59pmoreau: https://github.com/KhronosGroup/SPIRV-LLVM-Translator/pull/144 :-)
13:20imirkin: karolherbst: ah, could be
17:15rhysperry111: Hi guys, I’ve been reading up on nouveau recently and just wanted to know, has Nvidia released the firmware for it’s pascal cards yet? It took them two years for the previous generation and it has been 4 years since pascal was released
17:17Lyude: rhysperry111: they released firmware for gr and display related stuff, and did so for display on turing (can't remember if they got us gr firmware or not). We don't have any PMU firmware for any GPUs though starting with maxwell 2, which means anything maxwell 2 or later (with the exception of maxwell 2 laptops) can't be reclocked, which has a hugely negative impact on performance
17:17Lyude: and we've been promised said firmware for a while now
17:17Lyude: like, a while == years
17:18rhysperry111: So is the lack of power management the reason that my iGPU outperforms my 1050 TI?
17:18rhysperry111: Thanks for clarifying
17:23imirkin: Lyude: display firmware?
17:23imirkin: [technically we do have PMU firmware, but just not PMU firmware which enables management of power]
18:38pmoreau: Hahahaha, spent some time writing a script for diffing SPIR-V binaries and then doing some plumbing in Git so that it would automaticaly be used on git diff, and then I found out I could just set textconv to spirv-dis for *.spv files. :-D
18:41pmoreau: But it will be quite handy, when converting all the OpenCL-CTS binaries to use Input for BuiltIn and check that I only modified those parts.
18:42karolherbst: pmoreau: is there anything left to do on https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/2078 or should I go ahead and review it?
18:47pmoreau: I’m just going to add a comment for one of the functions, but otherwise everything is ready to go.
18:50pmoreau: karolherbst: Pushed with the change. Regarding the remaining opened thread on that MR, I’ll create an issue out of it once the MR is merged.
20:10Lyude: imirkin: mh-misspoke
20:25Lyude: skeggsb: jfyi - the new version of the nouveau crc series I'm about to send out in a few minutes has some changes on the nouveau bits, but you probably won't need to re-review fully it as it's just some style fixes from checkpatch. I'll make sure to list the changes I made in the patch as well
20:25Lyude: and hopefully this should be nearing the last respin, now that I've got all of the latency questions with the vblank worker stuff figured out