06:03jjsa: karolherbst Hello are you here?
07:09karolherbst: jjsa: what's up?
07:46jjsa: Karolherbst I have installed the kernel an performed a short test. all seem to work well and as expected.
07:46pmoreau: One thing that needs to be updated, is the versions we require for the dependencies. I think 0.2.1 is way too old for SPIRV-LLVM-Translator, and might even be the old API so it won’t even compile.
07:46pmoreau: I’ll double check that
07:50karolherbst: jjsa: cool
07:50karolherbst: jjsa: but do you know if with the old kernel nouveau.runpm=0 already fixed it or not?
07:50karolherbst: just so that I know which bug gets fixed
07:51karolherbst: pmoreau: sounds good
07:52pmoreau: Looking good otherwise
07:55jjsa: Karolherbst I have performed short tests with nouveau.runpm=0 and nouveau.runpm=1 in both case no problem. Actually the notebook run with nouveau.runpm=1. The powermanagement don't probably not work well, the fan ist in action.
07:57karolherbst: jjsa: I meant with the original fedora kernel
07:58karolherbst: jjsa: if runpm works, you should see a significant reduction in battery drain. Usually the fan is shared across the CPU and GPU and some firmwares keep the fans spinning quite fast overall...
07:58karolherbst: anyway, battery drain is the most reliable value
07:58jjsa: With runpm=0 this waht better but the nouveau driver crashed and it what not possible tu use it.
07:59karolherbst: okay, so it was runpm afterall. Good to know
08:00karolherbst: mhh or secboot
08:00karolherbst: either way
08:00karolherbst: glad to hear it is working with the patches
08:02pmoreau: karolherbst: How much do we care about Fedora 29? I would like to bump SPIRV-Tools to 2019.4, cause it has a lot of improvements to the validator (especially for OpenCL), and a fix for linking that we absolutely need when two input files share array types.
08:02karolherbst: not at all
08:03pmoreau: Okay, good
08:03karolherbst: f(N-2) is immediatly end of life after fN got released
08:03karolherbst: and f31 is quite near
08:04karolherbst: mhh, one month is the migration period actually
08:05pmoreau: (F29 is on 2018.4, but F30 and later are already using 2019.4 which is the latest version available at the moment.)
08:05jjsa: Karolherbst On the original kernel the system crashed totally. With runpm=0 the nouveau driver crashed later so that I had the crash report reported via dmesg. The boot time increase dramatically if the crash happen. Wiith the test kernel I had no seen any problems.
08:06karolherbst: yeah.. secbooting the GPU engines is painful... we still didn't figure out all the issues
08:06karolherbst: but at least we have a hopefully reliable workaround for the runpm issues now
08:07karolherbst: I also have a secboot workaround which works on my laptop, but... oh well
08:10jjsa: Karolherbst, I have just installed lm_sensors and hat a look to the gpu temperatur. There is anything wrong the reported temperatur ist 511 grad celsius.
08:10karolherbst: jjsa: ohh, that patch to fix that isn't upstream yet
08:10karolherbst: jjsa: 511 is -1 in 9 bits
08:11karolherbst: it should look like this: "temp1: N/A"
08:11karolherbst: but that gets fixed later
08:11karolherbst: well, it is upstream, but not in 5.2
08:12jjsa: Karolherbst this is a minor problem,
08:12karolherbst: could be promotoed for backporting... oh well
08:13karolherbst: there is some process in the kernel of doing it, but usually that already happens automatically for some reason so I never bothered
08:18pmoreau: Arffff, the exported version for SPIRV-LLVM-Translator is still 0.2.1, regardless of the API changes and what not. --"
08:19pmoreau:waves an angry fist
08:21pmoreau: karolherbst: Okay, feel free to merge and we will bump the dependencies later, once we get that mess sorted out.
08:27karolherbst: pmoreau: cool :) just open a bug for that :D
08:28pmoreau: Will do
08:56pmoreau: \o/ Party time! \o/ Finally merged! Thanks for picking up the work and all your hard work on it, karolherbst!!
12:46karolherbst: pmoreau: we also have to land the SPIRV frontend stuff as well
12:46karolherbst: mind rebasing the patches?
13:41pmoreau: karolherbst: Good point, I’ll do that.
13:41pmoreau: And open an MR.
13:43karolherbst: nice, my SVM branch still builds :)
13:44karolherbst: will check the patches and open a new MR as well
13:44karolherbst: and I think dave also wants to open some MRs :D
13:46pmoreau: Sounds good :-)
13:47pmoreau: Not sure what I will look into when I have some time again. Maybe Tesla support?
13:48karolherbst: sounds good
13:49karolherbst: pmoreau: https://gitlab.freedesktop.org/mesa/mesa/merge_requests/2076
13:50karolherbst: yeah, it should implement everything I think...
13:51karolherbst: and I found cl_arm_shared_virtual_memory
13:51karolherbst: which "backported" the 2.0 SVM functions to 1.2
13:51karolherbst: so everything except clEnqueueSVMMigrateMem
15:30pmoreau: karolherbst: https://gitlab.freedesktop.org/mesa/mesa/merge_requests/2078
15:31pmoreau: If you could add the “clover” label, that would be great!
15:31pmoreau: I should ask for those rights at least.
15:31karolherbst: do you have 25 commits in mesa? :D
15:32karolherbst: "43 Pierre Moreau"
15:33pmoreau: Apparently I do :-D
15:34karolherbst: ohh, I am close to 200 .. :D nice
15:35karolherbst: anyway, I added the label