00:00 Ako1: cat /proc/asound/card0/eld#4.0 monitor_present 1 eld_valid 1 monitor_name  connection_type HDMI eld_version [0x2] CEA-861D or below edid_version [0x0] no CEA EDID Timing Extension block present manufacture_id 0x1010 product_id 0x1010 port_id 0x1010101010101010 support_hdcp 0 support_ai 0 audio_sync_delay 32 speakers [0x10] RC sad_count 1 sad0_coding_type [0x0] undefined sad0_channels 1 sad0_rates
00:00 imirkin_: pastebin.
00:00 imirkin_: anyways
00:01 imirkin_: the 0x10 thing is familiar
00:01 imirkin_: that's the bug that nyef` fixed
00:01 imirkin_: you need a newer kernel.
00:01 nyef`: Yeah, that looks familiar alright.
00:02 nyef`: Evaluating ?P in emacs says 0x50, which would match that monitor name.
00:02 Ako1: ok. Thanks! I'll try to upgrade to new kernel
00:03 nyef`: It's a single-character fix if you want to patch your current kernel. drivers/gpu/drm/nouveau/nvkm/something/disp/hdagt215.c or something like that, there's an [0] that obviously should have been an [i].
00:04 imirkin_: https://git.kernel.org/cgit/linux/kernel/git/stable/linux-stable.git/commit/?id=2a3797ef69c527dc70df98485da9ab2e9d073edb
00:04 imirkin_: something like that? :)
00:05 nyef`: That'd be it.
00:07 Ako1: hmm... i'm afraid i'm not that advanced to patch and build a new kernel... but i'm sure i can install new version of kernel after some howto reading :)
00:07 nyef`: If you can build a new kernel, applying the patch is easy enough.
00:08 nyef`: Configuring a new kernel tree is a complete bloody nuisance, though. /-:
00:08 imirkin_: nyef`: how are your 3d adventures?
00:08 airlied: make localmodconfig
00:10 nyef`: imirkin_: On my todo list for last week and this, but bumped by higher-priority stuff like hacking my Forth implementation and knitting. I have the requisite hardware with me, for all four affected DISP types, plus a windows box and two kinds of IR emitter, at least.
00:11 imirkin_: cool. was just curious :)
00:14 Ako1: btw imirkin_ how did you translate 4.4.0-53 to 4.4.30?
00:14 imirkin_: Ako1: http://changelogs.ubuntu.com/changelogs/pool/main/l/linux-lts-xenial/linux-lts-xenial_4.4.0-64.85~14.04.1/changelog
00:14 imirkin_: search for 4.4.0-53
00:14 imirkin_: then scroll down until you see a "updated to v4.4.x" note
00:15 airlied: imirkin_: https://people.freedesktop.org/~airlied/piglit/cts-nouveau-open/problems.html
00:15 imirkin_: airlied: nice. 75 failed/crashed tests? not horrible...
00:16 Ako1: I have installed 4.4.0-64.85. so tomorrow will try again
00:16 airlied: doesn't look too bad, I'd have to run on skylake to get a good comparison
00:16 airlied: I can definitely say the arrays_of_arrays_gl fails are compiler ones
00:16 imirkin_: Ako1: unfortunately that's still only v4.4.44
00:16 imirkin_: whereas the fix only came into v4.4.48
00:17 imirkin_: airlied: btw, were you at the tip of VK-GL-CTS?
00:17 Ako1: ups... that is the latest i have in my software package manager
00:18 imirkin_: Ako1: yeah, that's the latest released by ubuntu
00:18 imirkin_: but unfortunately not the latest :)
00:18 imirkin_: the latest of the LTS series being 4.4.52
00:18 imirkin_: they'll catch up eventually.
00:19 airlied: imirkin_: yes
00:20 imirkin_: shader_integer_mix -- that's missing the extension enable iirc
00:21 imirkin_: the copy_image one's a bitch. might be the thing that finally makes me disable rgba4 support
00:29 airlied: I'll see if I can convince my skl to run it
00:34 Ako: it's me again :)
00:35 Ako: upgraded to 4.4.52-040452-generic
00:35 imirkin_: cool. that should have working hdmi audio... hopefully.
00:36 Ako: nope :( here is the output from cat eld
00:38 Ako: cat /proc/asound/card0/eld#5.0 monitor_present 1 eld_valid 1 monitor_name SAMSUNG connection_type HDMI eld_version [0x2] CEA-861D or below edid_version [0x3] CEA-861-B, C or D manufacture_id 0x2d4c product_id 0xc44 port_id 0x0 support_hdcp 0 support_ai 1 audio_sync_delay 0 speakers [0x1] FL/FR sad_count 3 sad0_coding_type [0x1] LPCM sad0_channels 2 sad0_rates [0xe0] 32000 44100 48000 sad0_bits [0xe0000] 16 20 24 sad1_co
00:38 imirkin_: ok, well that's much better.
00:38 Lyude: huh, this is definitely the first codebase I've ever seen that uses 3 space indenting
00:39 imirkin_: i suspect the new issue is that you're not outputting to the proper device? e.g. going to your system audio rather than hdmi audio?
00:39 imirkin_: Lyude: yeah, it's weird. you get used to it though :)
00:39 Ako: tried sound tests on both hdmi outputs
00:40 imirkin_: boooo
00:40 imirkin_: BOO!
00:40 imirkin_: nyef`: any suggestions?
00:43 imirkin_: Ako: could be that the 320M requires additional help to get hdmi audio going :( unfortunately the MCP89 chips were only put into macs...
00:44 Ako: i have mac mini 2010
00:49 Ako: and with nvidia driver i can get sound via hdmi. unfortunately, with nvidia driver i cannot get stable detection of connected monitors. with nouveau driver video is stable and autodetection works fine but there is no audio :(
00:50 imirkin_: can't win 'em all
01:21 nyef`: I'm back. Got caught up talking with some people.
01:21 nyef`: Mac Mini 2010 should *definitely* work with HDMI audio.
01:22 nyef`: And I say that because that's the system that I ran into the problem on.
01:24 imirkin_: Ako: --^
01:24 imirkin_: so i guess either something else got fixed after 4.4.x that didn't get backported
01:24 imirkin_: or ... user error? dunno
01:25 nyef`: What was I working with, 4.9, one of the 4.10 rcs?
01:26 imirkin_: something like that
01:27 nyef`: I'm not up for doing anything involved tonight, but I have my mini 2010 with me, so I can try to set it up and do a build/test cycle on 4.4.52 or whatever it is tomorrow.
01:27 nyef`: ... at least, I think I have my mini with me.
02:27 Horizon_Brave: hi.
02:29 gnarface: hello
02:40 jrayhawk_: What information is needed for new chipsets? "Unknown chipset: NV124" "01:00.0 VGA compatible controller [0300]: NVIDIA Corporation GM204GLM [Quadro M3000M] [10de:13fa] (rev a1) (prog-if 00 [VGA controller]) Subsystem: Lenovo GM204GLM [Quadro M3000M] [17aa:222d]"
02:42 Lyude: imirkin_: what would the gles2_ver in src/mesa/main/extensions_table.h for this be, 31?
02:42 Lyude: https://www.khronos.org/registry/OpenGL/extensions/NV/NV_fill_rectangle.txt
02:46 imirkin: jrayhawk: where are you seeing that?
02:47 imirkin: Lyude: GLES2
02:47 Lyude: alright
02:50 imirkin: i know it talks about ES 3.1, but that's just there to confuse you
02:50 Lyude: i had this weird feeling it was something like that, lol
02:51 imirkin: and GL 2.0 i suspect
02:51 imirkin: i'd just put GLL and ES2
02:51 imirkin: er
02:51 imirkin: not GLL
02:51 imirkin: been a while since i looked at that file, hold on
02:51 imirkin: yeah. GLL + GLC
02:52 imirkin: jrayhawk: if you're seeing that in kernel, you need a newer kernel. if you're seeing that in the DDX, you can grab DDX from git or use the modesetting DDX
02:52 Lyude: GL is weird
02:53 imirkin: jrayhawk: if you want acceleration to work on that GPU, make sure you have the nvidia-supplied firmware, in linux-firmware.git (and been there a while)
02:53 imirkin: Lyude: indeed it is.
02:54 Horizon_Brave: Hey everyone...I have a question that I asked last night, but I like a fool, closed my browser before I saw a response... When the nouveau community is writing/constructing a driver for a new nVidia chip/card, do they have to proactively seek out nvidia to get them to help out?
02:55 imirkin: Horizon_Brave: there are channel logs... see topic
02:56 Horizon_Brave: What information from nVidia is usually asked for? and does the nouveau developers when creating the driver, only able to write it for one particular linux kernel? meaning they have to write a different driver per kernel release? Or generally will 1 driver work across multiple kernels?
02:56 imirkin: Lyude: i take it back. no ES support until NV_oolygon_mode is implemented.
02:57 imirkin: Lyude: so only desktop GL
02:58 imirkin: Horizon_Brave: the nouveau kernel driver is part of the linux kernel. any updates go through the regular linux merging process.
02:59 Horizon_Brave: wait... so nouveau is literally 1 driver/module? It's just compiled of a lot of drivers for different nVidia devices??
03:00 imirkin: Horizon_Brave: you might benefit from browsing the wiki, also in the channel topic
03:00 Horizon_Brave: >.>
03:02 Horizon_Brave: okay okay, I can take the hint...but can you answer, do you guys from nouveau usually have to chase down and really badger the people at nVidia to help out?
03:02 Horizon_Brave: or when they release something new, they sort of toss you some scraps without you asking?
03:03 airlied: not sure there is any one method
03:03 airlied: sometimes a, sometimes b
03:14 Horizon_Brave: heh
03:17 gnarface: i get the impresion it's usually the former, and usually they withold help
03:18 airlied: gnarface: some areas are worse than others
03:18 airlied: like power management generally seems to elicit silence
03:58 airlied: imirkin: compared to skl: https://people.freedesktop.org/~airlied/piglit/cts-nouveau-open/changes.html
03:58 airlied: prob want to look at regressions there
04:07 Horizon_Brave: airlied, was that meant for me? Not sure what I"m looking at..?
04:07 Horizon_Brave: relevant to my question?
04:09 airlied: Horizon_Brave: nope
04:26 Horizon_Brave: oh..
04:31 airlied: Horizon_Brave: there have been times where nouveau devs have gotten pre-launch hw access btw
04:39 imirkin: airlied: thanks
04:40 Horizon_Brave: airlied: hmm so what changed? Why are the folks are AMD more willing to play ball with their open source for the Radeon developers?
04:40 airlied: Horizon_Brave: different companies have different customers and priorities
04:42 gnarface: also it's a strategically sound move for AMD because the performance of their own drivers is widely known to be marginal
04:43 Horizon_Brave: airlied: hmm fair enough... do you personally develop? how many people usually work on creating a driver for a piece of nVidia hardware? roughly how long does it take?
04:44 airlied: Horizon_Brave: I don't do much on nouveau anymore, but nouveau has 1 person full time and a small group of interested parties and devs
04:45 Horizon_Brave: gnarface: you mean to say that because even their proprietary "official" drivers usually under perform compared to nVidia's proprietary drivers, so they make up for this by offering more open source help?
04:45 airlied: each generation is different, some take a lot of work, some take hardly any, some need nvidia help, it can takes days to months to years
04:46 Horizon_Brave: airlied: lol what?? 1 person??
04:47 airlied: Horizon_Brave: skeggsb is the only full time paid developer
04:47 gnarface: Horizon_Brave: to be clear, understand that nvidia basically uses ONE driver for all their cards - this is a bit different from AMD's historical approach
04:48 gnarface: Horizon_Brave: but yea, essentially you got my gist
04:49 gnarface: they *seem* to genuinely need the help
04:51 Horizon_Brave: what do you mean? who needs what help? and if nvidia generally uses 1 driver across multiple devices, do they just sort of "update" it when newer functionality as newer hardware has more features etc..?
04:52 gnarface: i meant AMD in that context. and yea, nvidia just keeps updating it for a span of generations of cards, slowly rolling the oldest ones off as newer ones are added
04:53 gnarface: they have a legacy driver for previous cards, but i think it also now has a limited span (don't quote me on that though) and is generally unmaintained
04:53 Horizon_Brave: airlied: and if this skeggsb guy is the only main developer..who does he work for exactly? who pays him?
04:53 airlied: Horizon_Brave: Red Hat
04:54 Horizon_Brave: oh wow, I didn't know Red Hat had nouveau as a branch to support...
04:56 airlied: we have a need to certify RHEL on nvidia gpus, which means no binary bits for us
04:58 Horizon_Brave: >.> ...I would ask why...but I don't want to keep digging a deeper and deeper hole for you guys to answer more and more lol...
04:59 airlied: RH doesn't ship closed source components
04:59 airlied: so we have hw partners who need to certify their platforms
05:03 Horizon_Brave: ahh
05:03 Horizon_Brave: so then what about other distro's like Debian stable? they don't used closed source either
05:03 Horizon_Brave: why specifically redhat?
05:04 gnarface: debian just doesn't ship it, they do have it available in the optional repos
05:04 gnarface: ubuntu does ship it
05:04 gnarface: (afaik)
05:04 Horizon_Brave: oh your right
05:07 Horizon_Brave: hey, thanks in advance guys, I appreciate your time with all of this...
05:08 Horizon_Brave: patience, more like it
05:11 Horizon_Brave: is there any priority when developing drivers? if nVidia releases a bunch of cards at once (not sure if that even happens..) who determines which take priority and which ones should just be skipped ?
05:11 gnarface: they determine that themselves
05:12 gnarface: they don't keep it a secret which ones, either. you can track the adding/removing of support for various lines in their published changelog
05:12 airlied: usually the hw they release is pretty similiar
05:12 Horizon_Brave: when you say they, you mean the nouveau team? this skeggs guy and the part timers?
05:13 gnarface: oh i thought you were talking about nvidia's official drivers sorry
05:13 airlied: Horizon_Brave: lately the project relies on nvidia to release some binary firmware modules
05:13 airlied: when those get released to the project, determines when hw gets support a lot more often than not
13:53 dboyan: imirkin_: Are there really instruction like "cvt f64 $r8d f32 0x3f800000" on gk110?
13:53 dboyan: I was getting No match error on envyas
13:53 imirkin_: pretty sure i make use of precisely that instruction at some point
13:53 imirkin_: there might have to be a "rn" somewhere
13:55 imirkin_: looks like it's missing in gk110.c
13:55 imirkin_: but the op definitely exists
13:55 dboyan: any way to find it out?
13:55 imirkin_: yeah hold on
13:56 imirkin_: gr, i don't have my op lists on me
13:56 imirkin_: i fuzzed nvdisasm and got all the major opcodes
13:56 imirkin_: should be easy, hold on.
13:56 imirkin_: oh. i don't have nvdisasm here. i'll write the script for you though.
13:57 dboyan: okay, I have nvdisasm here
13:57 imirkin_: yeah, give me a minute
13:57 dboyan: okay
13:59 imirkin_: ok so something like this:
14:00 imirkin_: (for i in `seq 0 4095`; do printf "1 %03x00000\n" $i | perl -ane 'foreach (@F) { print pack "I", hex($_) }' > tt; nvdisasm -b SM35 tt; done) > gk110-1.txt
14:00 imirkin_: nvdisasm used to include the binary encoding as well. if it doesn't, there might be another flag you can pass in.
14:00 imirkin_: and then you're looking for a F2F op
14:01 imirkin_: 1 in the LSB is for immediate ops, so the F2F should be enough to find the F64 variant
14:04 dboyan: yeah, should add -hex for instruction encoding output
14:04 imirkin_: ok, it used to do that by default "back in the day" :)
14:08 dboyan: imirkin_: I was only getting things like "F2F.INVALIDFPDEST0.INVALIDFPSRC0"
14:09 imirkin_: great
14:09 imirkin_: which one is that?
14:09 imirkin_: (pastebin all of the F2F ones)
14:10 dboyan: http://pastebin.com/CBFd0Uj2
14:10 dboyan: The first form is the basic one, i think
14:11 imirkin_: oh yeah. so the op is 0xc54
14:11 imirkin_: and that 0x080 bit is the neg bit on the immediate
14:12 imirkin_: ok so now try....
14:12 imirkin_: (sec)
14:13 imirkin_: echo 2c01 c5400000 | perl -ane 'foreach (@F) { print pack "I", hex($_) }' > tt; nvdisasm -b SM35 tt
14:14 dboyan: @P0 F2F.F64.F32 R0, 0;
14:14 dboyan: wow
14:14 imirkin_: =]
14:14 imirkin_: now take all the N("cvt") lines
14:14 imirkin_: and copy them into tabi
14:14 imirkin_: and switch 02 to 01 and 254 to c54
14:15 imirkin_: and replace T(is2) with FIMM
14:15 imirkin_: or ... something
14:15 imirkin_: no wait, probably LIMM
14:15 imirkin_: hm. not sure.
14:15 imirkin_: will need some investigation
14:16 imirkin_: probably FIMM/DIMM/I3BIMM depending on the src type
14:23 imirkin_: anyways, that little trick is nice when you're searching for instructions... ops may either have 0 1 or 2 in the LSB 2 bits, which yield different instruction types. that's what e.g. emitForm_21() is about - it emits either a 2 or a 1-form op depending on the arg types
14:23 imirkin_: diff gpu ISAs end up with different structures, but something along these lines often appears
14:23 imirkin_: otoh, i think SM50 is just one continuous op space
14:32 dboyan: btw, I mentioned two week ago about a envydis issue, and you were apparently too busy for that.
14:32 imirkin_: what was the issue?
14:32 dboyan: imirkin_, It's about the first entry of tabus64_28
14:32 imirkin_: oh right yeah
14:33 imirkin_: welp, that SHOULD work
14:33 imirkin_: what's the issue, precisely?
14:34 dboyan: envydis will abort on instructions like 029ffc1a dfc02800
14:34 dboyan: on gk110
14:35 dboyan: seems that it is expecting a name in the end
14:38 imirkin_: hmmm
14:38 imirkin_: i guess removing that line should "fix" it
14:39 imirkin_: we use that pattern in lots of places though
14:40 imirkin_: mwk: any idea why that fails here? a line like { 0x0000000000000000ull, 0x0000030000000000ull }, hits an abort in core-dis.c:428
14:41 dboyan: well, that's what the blob was generating, completely no idea how it works
14:41 imirkin_: dboyan: blob generated an op like that?
14:41 dboyan: yeah
14:41 imirkin_: huh. i wonder what it does.
14:41 dboyan: 029ffc1a dfc02800: SHF.L R6, RZ, R5, R10;
14:42 imirkin_: i wonder if it gets the upper word of the shift.
14:42 dboyan: Especially mysterious when compared with the following:
14:42 imirkin_: (or lower in case of rshift)
14:42 dboyan: 029ffc1a dfc02a00: SHF.L.U64 R6, RZ, R5, R10;
14:42 dboyan: 029ffc1a dfc02b00: SHF.L.S64 R6, RZ, R5, R10;
14:48 dboyan: imirkin_, what exactly is the imm form I need for cvt f64 -> f32?
14:48 dboyan: A hint is 2c01 c5400200 : @P0 F2F.F64.F32 R0, 2;
14:48 imirkin_: FIMM probably
14:48 dboyan: okay
14:49 imirkin_: (based on logic rather than actually checking)
14:49 imirkin_: oh - for an added bonus - nvdisasm is sometimes wrong ;)
14:49 imirkin_: but that's very rare, and usually only on very weird things
14:50 dboyan: I did remember to see segfaults when running ptxgen though
14:51 imirkin_: right. sometimes nvdisasm also segfaults. it's a nice tool :)
15:02 imirkin_: dboyan: so did you work out what kind of immediate it is?
15:04 dboyan: imirkin_: wait a few minutes
15:11 dboyan: imirkin_: it is FIMM
15:11 imirkin_: cool. that's in line with what i expected.
15:20 imirkin_: dboyan: which means that the others are probably DIMM and I3BIMM
15:20 imirkin_: depending on the source datatype
15:20 imirkin_: the various modifiers have likely moved around
15:20 imirkin_: usually to right above the imm range.
15:23 dboyan: you mean other forms of cvt in tabi?
15:23 imirkin_: yeah
15:24 imirkin_: just copy all the stuff from tabm
15:24 imirkin_: and replace 254 with c54
15:24 imirkin_: 2 with 1
15:24 imirkin_: and then adjust
15:25 dboyan: will try that tomorrow, maybe. I don't have effort enough before sleep :)
15:25 imirkin_: hehe, no rush
15:26 imirkin_: i like the "cvt f32 -> f64" thing as a nice way to save an op
15:26 imirkin_: not that it matters in practice
16:43 billard: hello i'm new with IRC, and would some help for a NVIDIA Corporation GF108GLM [Quadro 1000M]
16:43 imirkin_: ask and ye shall receive
16:45 billard: ok i have install on Debian jessie Nouveau drivers, after i make xorg -configure and add a device section. When i try this Xorg.conf file i have NO DEVICE DETECT
16:45 imirkin_: i'd recommend clearing out any xorg.conf file that you created
16:46 imirkin_: things should Just Work (tm) by default
16:47 billard: I'am on a lenovo W520 with an inte card onboard and actually i use the intel drivers. Hox try Xorg with the nouveau drivers without lost my config
16:47 imirkin_: the screen should be driven by intel then, not nouveau.
16:48 imirkin_: either way, things should get automagically detected if you don't have an xorg.conf
16:49 billard: yes but finally i would use the external VGA, that actually does'nt work.
16:50 imirkin_: see https://nouveau.freedesktop.org/wiki/Optimus/ - "using outputs on discrete GPU"
16:50 billard: for the moment i don't have Xorgf.conf in /etc/X11
16:51 imirkin_: billard: pastebin xorg log, just to double-check
16:51 billard: ok thank's
17:03 karolherbst: imirkin_: I got the feral key :)
17:03 karolherbst: and it seems like you have one as well?
17:04 imirkin_: yes
17:04 imirkin_: but my desktop is still in a cardboard box. i really should unpack at some point.
17:04 karolherbst: ahh I see
17:04 karolherbst: so you wouldn't know if you have access to the hitmanpro game yet, but most likely you have
18:05 mwk: imirkin_: that's because you're using a subinstruction without an instrution name
18:06 imirkin_: mwk: we do that a lot though...
18:06 mwk: that's wrong
18:06 mwk: SESTART ... SEEND always needs to start with some name
18:06 imirkin_: so ... what should we do? N("") was bad too i thought
18:06 mwk: you can't just have ($a $b)
18:06 imirkin_: wait, it's not in a SESTART/SEEND thing
18:07 mwk: what is that anyway?
18:07 imirkin_: shf.l
18:07 imirkin_: https://github.com/envytools/envytools/blob/master/envydis/gk110.c#L924
18:07 imirkin_: oh shit
18:07 imirkin_: it is in a SESTART
18:07 imirkin_: who put it there! gr.
18:08 imirkin_: i blame my senility.
18:08 mwk: there's lots of it in there...
18:08 imirkin_: senility?
18:08 mwk: SESTART + us64_28
18:08 imirkin_: i know :)
18:08 imirkin_: i just looked
18:18 tarragon: hei
18:18 tarragon: can 4.9 force the card into suspend mode?
18:19 tarragon: it's stuck at 100% full throttle ALL the time.
18:19 tarragon: for a few terminals.
18:21 imirkin_: you mean the fan?
18:22 tarragon: fan, megahurtz, ram clockings etc.
18:22 imirkin_: what GPU?
18:23 imirkin_: mwk: so ... do we have an answer to that? just move the us64_28 to the end of the group?
18:23 tarragon: hold on
18:23 imirkin_: tarragon: lspci -nn -d 10de:
18:24 tarragon: gk104
18:24 imirkin_: that should be able to clock lower most of the time, i think. although it should boot to low clocks.
18:24 imirkin_: are you sure that's your issue?
18:25 tarragon: imirkin_: ah alright, but is automatically without kernel command line option and such?
18:25 imirkin_: tarragon: well, we don't touch clocks
18:25 tarragon: imirkin_: just asking since >4.8 implemented some re-clocking.
18:26 mwk: imirkin_: why is it a group in the first place?
18:26 imirkin_: kepler chips boot to min perf level i thought
18:26 imirkin_: mwk: coz it's a 64-bit quantity
18:26 mwk: so what happens when the field is 00?
18:26 imirkin_: mwk: search me
18:26 mwk: *sigh*
18:26 mwk: call it N("b64") or somethibng
18:26 mwk: a subexpression is supposed to have a name
18:26 imirkin_: allegedly the blob generates such code occasionally
18:28 tarragon: imirkin_: do I need pstate option ?
18:28 tarragon: I think I am using it.
18:28 imirkin_: no
18:28 imirkin_: if you are, then it will clock by default
18:29 imirkin_: tarragon: cat /sys/kernel/debug/dri/0/pstate
18:30 tarragon: I see a bunch of frequencies
18:31 imirkin_: pastebin
18:31 karolherbst: imirkin_: was there something special you had to do after activating the key?
18:31 tarragon: it's a different machine :/
18:32 imirkin_: tarragon: provide the last line
18:32 imirkin_: the one that says AC: bla
18:32 tarragon: AC: core 324 MHz memory 648 MHz
18:33 imirkin_: you're at your lowest clock speed
18:33 tarragon: imirkin_: excellent :D thanks
18:42 karolherbst: tarragon: you could enable clock gating
18:42 karolherbst: ohh, you mean your fan is at 100%
18:42 karolherbst: yeah, there might be a bug within nouveau regarding fan handling
18:42 tarragon: yeah, they're loud.
18:42 karolherbst: okay, so yeah, it
18:42 karolherbst: 's the fan handling
18:43 karolherbst: if you execute "sensors" you should see that your GPU is quite cool
18:43 tarragon: karolherbst: no problem, I though it might've been an option that overlooked in kernel 4.9
18:43 karolherbst: you should be able to manucally control your fan though I think
18:43 karolherbst: *manually
18:43 karolherbst: mupuf: ^^
18:43 tarragon: I was concerned the 100% full throttle fans might reduce the bearing balls lifespan.
18:43 karolherbst: tarragon: there are some fan related fixes though
18:45 tarragon: karolherbst: i'll pass the manual way, I don't want to get my card fried accidentally if I forget to turn them up during gaming.
18:45 karolherbst: you won
18:45 karolherbst: 't
18:45 karolherbst: the GPU shuts itself off if it gets too hot
18:46 karolherbst: but yeah, you might want to still increase the fan speed before using something demanding :)
18:48 mupuf: karolherbst: probably is the same bug I have been trying to fix
18:48 mupuf: changing the speed by hand won't do anything
18:49 karolherbst: k
18:49 karolherbst: yeah well, then there is another one with the same problem
18:50 karolherbst: any good enough workaround possible?
18:50 imirkin_: mupuf: that was on fermi
18:50 imirkin_: mupuf: while this is kepler
18:50 mupuf: imirkin_: I doubt that this is limited to fermi
18:53 imirkin_: ah
19:00 Lyude: imirkin_: you said we've already got the stuff for NV_fill_rectangle in gallium right?
19:00 imirkin_: no
19:00 imirkin_: it was already there for EXT_polygon_offset_clamp, so you wouldn't see those bits in my patches
19:01 Lyude: ah, right
19:03 Lyude: have any examples of a simple gallium commit you would recommend I look at btw?
19:04 imirkin_: let's see....
19:04 imirkin_: 59a692916ca251db995050f7fc0bb7b4e6e4780b is an example of adding something to the rast state
19:05 imirkin_: ooooh
19:05 imirkin_: look at my EXT_window_rectangles series
19:05 imirkin_: that should cover everything
19:05 Lyude: imirkin_: gotcha, thanks for the info!
19:06 imirkin_: Lyude: https://patchwork.freedesktop.org/series/8590/
19:07 imirkin_: so there i create a whole new state for the window rects. you wouldn't do that - you'd modify pipe_rasterizer_state.
19:08 imirkin_: but the other bits apply
20:14 Lyude: imirkin_, skeggsb: I'm thinking I should bump the release version for the DDX since that commit ^ is required for the ddx to not be mostly broken with atomic modesetting on nouveau, do you have anything else you want in before I do that?
20:15 imirkin_: sgtm
20:15 imirkin_: i've been too chicken to do a release myself
20:15 imirkin_: with the GM20x support
20:15 imirkin_: i guess dropping GP10x support would be nice but ... meh. skeggsb hates the idea, and i probably won't have a card to play with for another year.
20:15 Lyude: We're still doing support for new nvidia GPUs in the ddx?
20:15 imirkin_: define "we"
20:15 imirkin_: i am ;)
20:16 Lyude: ahh
20:16 imirkin_: skeggsb isn't
20:16 imirkin_: i think using glamor is a horrible idea. skeggsb thinks it's a great idea.
20:16 Lyude: i'm on the side of modesetting/just move to wayland myself, but I understand
20:17 imirkin_: i'm on the side of "simple, never-changing code with trivial error handling complexity = better for an X server that shouldn't crash"
20:17 imirkin_: anyways, i think we're at peace with each others' opinions
20:19 imirkin_: "uncrashable" is a word that gets thrown around a lot, but xf86-video-nouveau is pretty solid
20:23 hakzsam: imirkin_: maxwell sched codes maybe?
20:23 imirkin_: that'd be nice, but i'm not about to do it
20:23 imirkin_: and it didn't sound like you had the time for it either
20:24 hakzsam: not much yeah
20:24 airlied: should just use glamor :-P
20:25 imirkin_: i realize it's pretty much me against the world
20:25 imirkin_: but i don't really care.
20:25 imirkin_: the world is wrong :p
20:26 nyef`: ... Why am I thinking of an old Duck Tales episode involving some sort of craft known as "The Uncrashable Hindentanic"?
20:26 imirkin_: https://s-media-cache-ak0.pinimg.com/736x/5f/e5/b0/5fe5b02a414796ca27bf7d7e86ee62a8.jpg
20:27 imirkin_: heh. i remember watching Duck Tales translated into Russian. don't think i've ever seen it in English
20:28 mlankhorst: dutch here
20:29 imirkin_: it was among one of the first foreign tv shows/cartoons available in the late 80's
20:29 mlankhorst: though I can't really recall watching it, apart from the intro song
20:30 imirkin_: with, as i recall, some extremely misplaced ad about it being translated/made available by Johnson & Johnson, even though the viewing public didn't have a clue as to what that was.
21:21 Ako: Hi! I have a question. Has anyone been able to use HDMI Audio on MacMini 2010 with nouveau drivers?
21:22 imirkin_: nyef` has, as per his comments yesterday
21:23 Ako: Ya. I do remember. however I still can't set it up properly. Even after upgrading kernel to 4.4.52
21:23 imirkin_: any chance you can try a 4.10 kernel?
21:24 Ako: hmm... I can try. Which one?
21:24 imirkin_: 4.10.0 or 4.10.1 are fine.
21:39 karolherbst: imirkin_: ... the key works, but the game which is missing is hitman... :(
21:39 karolherbst: but I got most of the other stuff
21:39 Ako: Installed 4.10.1 - still no sound via HDMI
21:41 imirkin_: karolherbst: lol, seriously? you should ask him about it.
21:41 imirkin_: Ako: ok. well, either you or nyef` is lying.
21:42 imirkin_: Ako: confirm you're really running 4.10.1 -- run "uname -r"
21:43 Ako: I have tried to install ubuntu, linux mint, elementary os, fedora... the same result - no HDMI sound. Yes - I have checked uname -r
21:44 imirkin_: my money's on nyef` ... he seems like quite the trickster...
21:44 Ako: or he knows some other tricks
21:44 Ako: to make it work
21:44 imirkin_: Ako: maybe EFI vs BIOS boot?
21:44 imirkin_: (was that a thing for that mac mini?)
21:45 karolherbst: imirkin_: I already got an answer, it will be added later, cause it just got released
21:45 imirkin_: karolherbst: ah ok
21:45 karolherbst: imirkin_: sometime this week
21:45 imirkin_: i'm in no rush :)
21:45 karolherbst: it was just odd, cause I thought I didn't got the games
21:45 karolherbst: cause I owned around 50% already
21:45 imirkin_: karolherbst: let me know if you find anything noteworthy with the other games
21:45 karolherbst: and just checked the wrong ones :D
21:45 Ako: I doubt so... Actually when I install proprietary Nvidia driver I can make it work but only if I connect TV alone
21:45 karolherbst: imirkin_: I will test deus ex
21:45 imirkin_: Ako: well, the issue could be that we're missing some bit of init
21:46 imirkin_: Ako: that nvidia does and e.g. bios does but not efi.
21:46 karolherbst: ohh, and xcom 2 I guess
21:47 Ako: if I connect both DVI and TV - nvidia driver refuses to work
21:47 Ako: for nouveau it's opposite - I can get picture on both monitors but the sound on TV is missing
22:30 nyef`: Oh, right, I was going to set up my Mini today, wasn't I?
22:52 nyef`: Prepping a fresh 4.10.0 kernel build. Hardware has been a bit... finicky about booting, doesn't want to load any of the -rc kernels, but does load the older distro kernel.
23:40 nyef`: Ah! Blacklisted modular nouveau, and efifb not kicking in for whatever reason.
23:40 nyef`: Which would explain the working-to-some-extent keyboard.
23:42 karolherbst: imirkin_: we got access to hitman now :)
23:42 nyef`: And /proc/asound/card0/eld#5.0 on stock 4.10.0 recognizes the TV...
23:44 nyef`: ... And "aplay -D hw:0,8 test-file.wav" directs output such that the TV (attached to the HDMI port on the back of the machine) mediates the audio (tested by using the mute button on the TV remote).
23:45 nyef`: So: HDMI audio on a Mac Mini 2010 (MCP89), on 4.10.0, works for me.
23:47 hakzsam: karolherbst: but it's just plain broken, no?
23:49 karolherbst: hakzsam: we fixed it so that we can replay radeonsi traces without crashes
23:49 karolherbst: hakzsam: compute was broken ,)
23:50 karolherbst: hakzsam: https://cgit.freedesktop.org/mesa/mesa/commit/?id=8e6d67685e10b001e07f92a7a6aaff4fe987b6f2
23:50 hakzsam: yeah, imirkin_ just not fixed it correctly :)
23:50 karolherbst: will test the game after I am finish downloading it
23:51 hakzsam: presumably, it won't work because MT is still broken
23:51 karolherbst: yeah
23:51 karolherbst: mhh
23:51 karolherbst: maybe
23:51 karolherbst: but I never ran into MT issues for now
23:51 karolherbst: not really
23:51 hakzsam: skeggsb fixed it?
23:51 karolherbst: I have the advantage, that I don't run X on the nvidia gpu
23:52 karolherbst: no, just less contexts overal
23:52 karolherbst: l
23:52 karolherbst: don't have issues with MT piglit as well for like always
23:52 hakzsam: ah yeah, all the recovery fixes
23:52 karolherbst: not really
23:52 karolherbst: never hit them yet
23:52 hakzsam: it's easy :)
23:52 karolherbst: tell me how
23:53 hakzsam: just run arb_shader_image_load_store-atomcity
23:53 karolherbst: piglit?
23:53 hakzsam: yup
23:53 karolherbst: how often in parallel?
23:53 hakzsam: maxwell though
23:53 hakzsam: no, it's unrelated to MT but skeggsb improved the situation with his recovery work
23:54 karolherbst: I see
23:54 hakzsam: the test probably still fails for weird reasons thouhg
23:54 hakzsam: I never figured why
23:54 karolherbst: what is the best way to run it?
23:55 hakzsam: bin/arb_shader_image_load_store-atomicity -auto
23:55 karolherbst: k
23:55 karolherbst: it's good to have access to those games :) I like that. More publishers should be like this
23:55 hakzsam: yeah
23:55 hakzsam: and it's an awesome game
23:55 hakzsam: even for a non player like me
23:55 karolherbst: you should play life is strange
23:56 hakzsam: I have it, just no time:)
23:56 karolherbst: play it :p
23:56 karolherbst: then you will find time
23:56 hakzsam: hehe
23:56 hakzsam: I usually don't play, just replaying apitrace and launching benchmarks ;)
23:57 karolherbst: yeah, I figured
23:57 karolherbst: but I wouldn't say this, if I wouldn't be sure it is a really great game
23:57 hakzsam: I trust you
23:58 karolherbst: the test passes
23:58 hakzsam: chip
23:58 hakzsam: ?
23:58 karolherbst: gk104
23:58 hakzsam: expected.
23:58 karolherbst: ohhh wait
23:58 karolherbst: gk106
23:58 karolherbst: like there is a difference
23:58 karolherbst: :D
23:58 hakzsam: it only with maxwell
23:58 karolherbst: k
23:58 hakzsam: +fails
23:59 karolherbst: hakzsam: ohh wait, for you as well (and also pmoreau and RSpliet and imirkin_ and everybody else I forget to mention). I have no a place near the center of hamburg to live, so whenever you want to visit hamburg, you can stay at my place
23:59 hakzsam: cool, a new flat?