00:00karolherbst: mystified: do you have a /boot/syslinux.cfg file?
00:00karolherbst: or /etc/update-extlinux.conf
00:03karolherbst: imirkin: I get a "/home/karol/Dokumente/repos/mesa/build64/lib/gallium//nouveau_dri.so: error: symbol lookup error: undefined symbol: draw_gs_llvm_destroy_variant (fatal)"
00:03mystified: x58-al:/boot# ls
00:03mystified: System.map-grsec initramfs-grsec libutil.c32 vesamenu.c32
00:03mystified: boot ldlinux.c32 lost+found vmlinuz-grsec
00:03mystified: config-grsec ldlinux.sys mboot.c32
00:03mystified: extlinux.conf libcom32.c32 menu.c32
00:05karolherbst: mystified: I guess you have a "/etc/update-extlinux.conf" file then
00:05karolherbst: but I have _no_clue_ about how alpine linux does this stuff
00:06imirkin: karolherbst: sounds like you did something weird :p
00:06karolherbst: maybe there is a default_kernel_opts= with nomodeset
00:06karolherbst: and then you have to run update-extlinux ?
00:06imirkin: karolherbst: definitely not connected to my patches...
00:06karolherbst: imirkin: yeah
00:06karolherbst: imirkin: I guess I hit this issue? https://bugs.freedesktop.org/show_bug.cgi?id=99010
00:06karolherbst: I had that flag set
00:07karolherbst: or somethine related to that
00:08karolherbst: mystified: is/was there a nomodeset within the update-extlinux.conf file?
00:08karolherbst: line 15 )
00:09karolherbst: remove nomodeset
00:09karolherbst: and run update-extlinux
00:09imirkin: karolherbst: could be yea
00:09karolherbst: imirkin: at least glxinfo works now
00:10mystified: the whole line
00:10karolherbst: mystified: no, just nomodeset
00:10karolherbst: imirkin: [770/770] pass: 710, fail: 60
00:11imirkin: karolherbst: grrrrr. which ones fail?
00:11imirkin: karolherbst: piglit-summary.py int64 | grep fail
00:12mystified: so leave "quiet"
00:12karolherbst: mystified: yes
00:12imirkin: karolherbst: can you pull in the 2 piglit patches i sent
00:12mystified: that part done
00:12karolherbst: imirkin: do you have a public piglit repository?
00:12imirkin: karolherbst: top two - https://patchwork.freedesktop.org/project/piglit/list/?submitter=11061
00:13karolherbst: using git would be much easier :/
00:13mystified: reboot ?
00:14karolherbst: mystified: run update-extlinux
00:16karolherbst: then reboot
00:16mystified: tHx ... u guys are Incred +++++1000
00:16karolherbst: imirkin: https://gist.github.com/karolherbst/9e2e1d74a4724f990589688cf7c82455
00:17imirkin: ok. something with lshift.
00:19OUTBURST: what goodies are in 4.8?
00:19imirkin: karolherbst: will get back to you.
00:22imirkin: karolherbst: i pushed an updated branch earlier... i don't think it should make a difference, but mind giving it a shot? we can focus on generated_tests/spec/arb_gpu_shader_int64/execution/built-in-functions/fs-op-lshift-uint64_t-uint.shader_test for now.
00:23karolherbst: I already updated except you updated again
00:23imirkin: karolherbst: if it still doesn't work, mind giving the NV50_PROG_DEBUG=1 output?
00:23karolherbst: HEAD is 31c38ca0bafa186666d3b16cf243b2dd549ac50d
00:23imirkin: yeah that's my HEAD on int64 too.
00:24mystified: back .. Good & bad news.
00:24mystified: bad news failed & and on diff distro
00:24karolherbst: imirkin: how can I run one test with the piglit tools? or do I have to use shader_runner for that?
00:26karolherbst: ohh, that's part of piglit as well
00:26karolherbst: silly me
00:27karolherbst: imirkin: https://gist.github.com/karolherbst/fe9b14671aa49b8314a7cf5315c80418
00:45imirkin: karolherbst: in NVC0LegalizeSSA::handleShift can you adjust the condition to instead be chipset < NVISA_GK20A_CHIPSET ?
00:45imirkin: hold on, let me just push an updated patch
00:47imirkin: karolherbst: pushed. please try again.
00:49karolherbst: imirkin: everything passes now
00:55imirkin: karolherbst: thanks! is that all the tests, or just that one?
00:55imirkin: (i mean all the int64 tests)
00:55karolherbst: all int64 ones
00:55imirkin: awesome, thanks for testing
00:59mystified: hey guys..
01:02karolherbst: mystified: we really can't help with distribution specific issues though
01:03mystified: i know. & gr8ly appreciate your help..
01:03mystified: But can I d=share with you what happend
01:04karolherbst: I will be gone any second now
01:05mystified: tx.. I really do apreciate the hrlp
09:31hakzsam: imirkin: nice
09:35milesrout: i have a 980 ti and a 770, I want to help with nouveau, where can I start?
10:17milesrout: shutdown now
10:17milesrout: oops haha
10:17milesrout: typed that with screen off
10:22mwk: that never ends well
10:22mwk: think of all the machines you could've been sshd to...
10:24mlankhorst: /amsg The system is going down for power off NOW!
16:01headless: imirkin_: hi. the issue turned out to be more complex than just passing nouveau.modeset=0
16:02headless: I'm not sure what the conditions are (cold boot?) but the mate-panel hang still happens
18:51kattana: milesrout: lemme see, I'd love some opnecl support!
18:53mupuf: kattana: ah ah
18:55kattana: mupuf: uh? tesseract needs it, darktable, etc. Its a reasonable goal.
18:55kattana: --> 10:35 < milesrout> ... I want to help with nouveau, where can I start?
18:56mupuf: I know, I know
18:56mupuf: just laughing because you are asking a newcomer to do this, and it is really non-trivial
18:56mupuf: pmoreau is working on opencl, on his spare time
21:37pmoreau: imirkin: Sorry, I have been busy over the week-end, and traveling some more tomorrow. Give me a few minutes to go through the logs :-)
21:39imirkin: pmoreau: no worries - i sent patches, which are mostly complete. i have updated versions in my int64 branch
21:40hakzsam: imirkin: what did you fix for the maxwell issue?
21:41imirkin: hakzsam: https://github.com/imirkin/mesa/commit/9de73c30e23fcd3e7566de7887583ba06c74d7a1
21:42hakzsam: does this fix all int64 fails?
21:42hakzsam: (on maxwell)
21:42hakzsam: my sched logic is not totally stupid :)
21:43imirkin: not totally :)
21:44pmoreau: hakzsam: Nope, a concert in Paris, and taking a few days to go back home. :-)
21:45pmoreau: imirkin: I’ll have a look at your branch. :-)
21:45hakzsam: pmoreau: cool
21:46pmoreau: I only had a look at CVT and MAD/MUL for S64/U64.
21:46pmoreau: I guess that MAD/MUL is taken care of before generating the TGSI?
22:36imirkin: pmoreau: i took your splitting pass into my series
22:37pmoreau: Yup, I saw that
22:37pmoreau: (After having written my comment)
22:38imirkin: worked like a charm
22:38pmoreau: Cool :-)
22:38imirkin: and i left int div/mod to the reader. for now it's lowered by the higher levels.
22:39pmoreau: How far did you went with the CVY ops?
22:39imirkin: not far at all... i just didn't use them - they don't work
22:39imirkin: i lower them at the from_tgsi stage
22:41pmoreau: So, each backend is expected to lower them?
22:44imirkin: oh, you mean a hypothetical from_spirv? yeah, dunno what the best thing to do is tbh.
22:45imirkin: it was easiest to do it there so that's what i did. i'm also not 100% happy with my min/max handling either.