03:12 HdkR: Oof, I booted Ubuntu on this laptop with an MX150 and the nouveau kernel driver was NOT happy with it
03:12 imirkin: which one's that? GM107?
03:12 imirkin: or did they make updated ones with GP108 or whatever/
03:13 HdkR: Wikipedia says GP108
03:13 imirkin: ah
03:13 HdkR: `02:00.0 3D controller: NVIDIA Corporation GP108M [GeForce MX150] (rev a1)`
03:13 imirkin: yeah, there are all sorts of assorted PM issues
03:14 imirkin: _DSM vs the new-and-improved-and-non-working thing which is the default
03:14 imirkin: [ah right. 940MX was the GM108 one.]
03:15 imirkin: HdkR: nouveau.modeset=0 should fix it right up
03:16 HdkR: `02:00.0 3D controller: NVIDIA Corporation GP108M [GeForce MX150] (rev a1)`;p
03:16 HdkR: ...
03:16 HdkR: https://hastebin.com/radudezova.sql Snipped kern.log for you to look at :)
03:17 HdkR: Getting used to these arrow key placements
03:17 imirkin: Dec 15 08:04:06 ryanh-Blade kernel: [ 3.383850] nouveau: detected PR support, will not use DSM
03:18 imirkin: nouveau 0000:02:00.0: bus: MMIO read of 00000000 FAULT at 6013d4 [ TIMEOUT ]
03:18 imirkin: ahahaha
03:18 imirkin: skeggsb: --^
03:18 imirkin: we've just been discussing how to get rid of that
03:18 imirkin: but that's not your problem
03:18 imirkin: your problem is that gr doesn't init properly
03:18 HdkR: I actually blacklisted Nouveau then installed the blob and disabled it since I won't use the thing often enough to matter :P
10:19 karolherbst: HdkR: uhh, mhh, I thought we fixed that :/ crap
10:21 karolherbst: HdkR: try this: https://github.com/karolherbst/nouveau/commit/ebfe26bbf0558b0972cdd70790a270a2f9c1bccc
12:26 karolherbst: imirkin: yep.. inverseCondCode is wrong for floats
12:26 karolherbst: set u32 %r26 neu f32 %r24 %r25 + slct u32 %r29 ne %r28 %r27 %r26 => slct ftz u32 %r29 equ f32 %r27 0.500000 %r24
12:26 karolherbst: equ is obviously wrong
12:28 karolherbst: wondering why there is a equ to begin with though... need to take a deeper look at the shader, but equ would be wrong in either case
12:30 karolherbst: okay.. load propagation is doing something
12:30 karolherbst: "slct ftz u32 %r29 neu f32 %r28 %r27 %r24" -> "slct ftz u32 %r29 equ f32 %r27 0.500000 %r24"
12:30 karolherbst: mhh, other constant would be a 0x3f800000
12:32 karolherbst: anyway, if you swap sources you still need to change NEU to EQ, not EQU
12:32 karolherbst: beause (!(a == 0.0)) == (a != 0.0)
12:33 karolherbst: a != 0.0 is a NEU 0.0, a == 0.0 is a EQ 0.0. and !a == 0.0 needs to be a NEU 0.0
12:44 karolherbst: imirkin: this is what would happen with my patches and codegen as it is today: https://gist.githubusercontent.com/karolherbst/806db612b79fe25b8993f6c3c821b7da/raw/d5f77906174f26994c96c0180ca8a633bab2f61e/gistfile1.txt
12:44 karolherbst: and I am sure the last step is wrong
12:47 karolherbst: uhm...
12:47 karolherbst: https://gist.githubusercontent.com/karolherbst/806db612b79fe25b8993f6c3c821b7da/raw/c3ddb7a93a61fa82db20b83958c50c70c455243c/gistfile1.txt
12:54 karolherbst: mhh a possible patch to fix it: https://github.com/karolherbst/mesa/commit/7079da0c4ff12aa35e29ad1b85cb5c9e26d69a89
19:39 HdkR: karolherbst: Interesting. I may be able to give it a try later
19:43 karolherbst: HdkR: I still have it for my gp107...
19:43 karolherbst: no idea what is going on though
19:44 karolherbst: kind of thought that gp108 firmware update helped, but aperantly there were still issues
19:46 imirkin: lots of people have PM issues
19:46 imirkin: with the new thing
19:46 imirkin: but yeah, could be the firmware thing - HdkR is your firmware up to date?
19:46 imirkin: (linux-firmwarE)
19:46 HdkR:checks
19:46 imirkin: it was long enough ago that i forgot about it
19:46 imirkin: but with an off-the-shelf distro... who knows
19:49 karolherbst: imirkin: I have the newest firmware and still issuses apperantly
19:49 HdkR: I have the GP108 firmware blobs from linux-firmware. Looks like there was only one commit for those
19:49 karolherbst: yeah, the gp108 firmwares are the update for all the other
19:49 imirkin: HdkR: they were updated iirc
19:49 karolherbst: *others
19:49 imirkin: oh right
19:49 imirkin: yeah, the pre-GP108 ones were the ones that were updated.
19:49 karolherbst: imirkin: saw my comment regarding inverseCondCode and floats?
19:50 imirkin: karolherbst: i did. requires thought. i hate thought.
19:50 imirkin: ;)
19:50 karolherbst: well
19:50 karolherbst: != is NEU for floats ;)
19:50 karolherbst: but yeah
19:50 karolherbst: it's weird
19:51 HdkR: imirkin: Ah, I see that now. Let me check
19:51 karolherbst: with the same logic you could argue that LT should get inverted to GEU if you for example switch the args of a slct
19:51 HdkR: Was looking at gr/ specifically, oops
19:51 HdkR: Yep, don't have that one
19:52 HdkR: https://git.kernel.org/pub/scm/linux/kernel/git/firmware/linux-firmware.git/commit/nvidia/gp108?id=85c5d90fc155d78531efa5d2b02e92aaef7e4b88
19:52 HdkR: Four months ago. Disappointed Ubuntu :P
19:52 HdkR: Three*
19:53 karolherbst: HdkR: doesn't affect you
19:53 karolherbst: you have a gp108 ;)
19:53 HdkR: It changed the blobs in the gp108 folder though
19:53 karolherbst: nope
19:53 karolherbst: HdkR: mv gp108/ gp102; ln -s gp108 gp102
19:53 HdkR: Oh, it's a link
19:54 karolherbst: yeah
19:54 HdkR: derp, alright