IRC Logs for #crux Monday, 2017-08-14

brian|lfssounds about right or chromium00:11
*** cippp has joined #crux00:13
jaegerI used to think llvm was a long one but after doing all those build plots firefox was much longer00:23
brian|lfsI can do firefox in 20 to 30 minutes chromium takes a good hour00:35
jaegerI'd like to do another set of build plots for firefox but the ryzen box crashes too much :P00:37
jaegerI guess I could script it to keep trying until it has X number of successful builds, hrmm00:38
*** j_v has quit IRC00:44
nullspoonHas anyone tried installing xclip lately? I rebuilt my system, to 0.13 had to be downloaded and I'm getting a 404 on the source.01:23
nullspoonThe source shows the file exists under the home directory of rehabdoll, under the crux.nu domain.01:24
*** Dell has joined #crux01:35
jaegerfrinnst: ^01:38
jaegerI think he changed the username so that particular path no longer exists01:38
jaegerwith that said, you can grab it from one of the distfiles mirrors01:39
*** pejman has quit IRC01:42
*** mavrick61 has quit IRC02:27
*** mavrick61 has joined #crux02:28
joacimhttps://i.imgur.com/J7JQE8v.jpg02:30
joacimgot my old windows 7 install working in qemu, with gpu passthrough02:30
joacimbut the performance is less than stellar02:30
cruxbot[compat-32.git/3.3]: mpg123-32: 1.25.5 -> 1.25.602:37
*** pejman has joined #crux02:40
*** pejman has quit IRC02:40
*** pejman has joined #crux02:40
Romsteron what hardware joacim ?02:48
joacimhttps://pcpartpicker.com/list/PqhwvV02:48
Romsteron the R9 for the VM?02:49
Romsterwhat is the linux side using for video? a second R9?02:50
joaciman rx 55002:50
Romsteri am guessing you got the most recent firmware and microcode02:50
joacimi yeah updated just a couple of days ago02:51
Romsteranything special to use vga pass through? i'd like to try it when i get myself a new pc.02:52
jaegeryou need hardware that supports IOMMU02:52
joacimcan be fiddly if you have two graphics cards that use the same drivers02:52
jaegeras well as a motherboard and BIOS/UEFI (UEFI recommended) which has properly-separated IOMMU groups02:53
joacimi "solved" that by reserving my graphics cards at boot, and keep amdgpu loading as a module02:53
Romsteri heard some boards even if they support IOMMU wont dedicate a PCIe slot for the VM02:53
joacimryzen computers had horrible iommu groups, everyone fell into one of two groups. but the update that released this summer solved much of that02:54
Romsteri can't think of any reson why the windows driver for video would be an issue. must be a qemu vfio issue?02:55
joacimwill probably need a kvm switch too, if you want to switch between controlling the host system and the guest02:55
Romsteri was reading about that02:55
Romsterby the time i can afford one i should be okay (i hope)02:55
joacimi just got one of those cheap usb only ones =)02:56
Romstereh a second keyboard and mouse and point them to qemu?02:57
joacimthat would work too02:57
Romsterbe like having two computers with 2 or more screens02:57
Romsterif you got the desk space02:57
joacimi want to use passthrough for games, so i dont have to have two sets of input devices on my desk anymore02:57
joacimbut i could just have two computers and a kvm switch02:58
Romsteri'm getting sick of dual booting and i want to do the same as you joacim02:58
joacimyeah thats why i ended up with two computers. got tired of how inconvenient dualbooting was02:58
joacimi often found myself wanting documents or something that was on my other partitions, stuff i couldnt access from windows02:59
Romsterwonder what hypervisor you are using03:00
jaegerMy old haswell setup did qemu vfio really well, haven't tried it on the skylake03:01
Romsteri was thinking of a NAS to solve that.03:01
jaegerand the ryzen box doesn't have room :)03:01
*** pejman has quit IRC03:01
RomsterThe long story is that according to Microsoft's documentation "On systems03:02
Romsterwhere the TSC is not suitable for timekeeping, Windows automatically selects03:02
Romstera platform counter (either the HPET timer or the ACPI PM timer) as the basis03:02
Romsterfor QPC." QPC = QueryPerformanceCounter() which is a windows api for getting03:02
Romstertiming info. Some redhat documentation say: "Windows 7 do not use the TSC as03:02
Romstera time source if the hypervisor-present bit is set". Instead if falls back on03:02
Romsteracpi_pm or hpet if hpet is enabled in the xml.03:02
joacimi dont know too much about these things, so i cant tel you which hypervisor i'm using =)03:02
RomsterThe hypervisor present bit is a fake cpuid flag qemu and other hypervisors injects to show the guest it's running under a hypervisor. This is different from the KVM signature that can be hidden with <kvm><hidden state='on'>. With the hypervisor flag disabled in libvirt xml windows 7 started using TSC as timing source for me.03:03
RomsterThe performance for those games are basically limited by how fast the host can do context switches.03:04
Romsterso i dunno might be wroth looking into.03:04
Romsterworth*03:05
joacimhttps://paste.pound-python.org/show/yWJ1WcXxvUrfCUGai5Yu/03:05
joacimthis is what im using to start qemu03:05
Romsterhmm ia m pretty sure this pc has VT-d, but i am limited to 8GB of ram so i wont even bother on this pc03:08
*** pejman has joined #crux03:09
*** pejman has quit IRC03:09
*** pejman has joined #crux03:09
Romsteri would assume you are using the ACPI timer and not HPET with that configuration03:09
*** pejman has quit IRC03:14
joacimi can at least use qemu to get updates for windows. i think crux should also deny updates to users of newer processors03:30
*** pejman has joined #crux03:48
*** pejman has quit IRC03:48
*** pejman has joined #crux03:48
*** SiFuh has joined #crux04:41
*** tilman has joined #crux04:44
*** BitPuffin|osx has quit IRC05:42
*** nwe has quit IRC05:56
*** nwe has joined #crux06:00
SiFuhpedja: Fri Jul 28 14:03 - 16:09 (17+02:05) & Thu Jul 13 00:53 - 14:01 (15+13:08)06:10
SiFuhI just checked the previous uptimes06:11
*** izdmar has joined #crux06:39
*** izdmar has quit IRC06:41
*** sonophilia has quit IRC06:50
pedjajoacim, you might try this https://scottlinux.com/2016/03/21/enable-hyper-v-enlightenments-in-kvm-for-better-windows-vm-performance/07:02
pedjatl;dr is to add 'hv_relaxed,hv_spinlocks=0x1fff,hv_vapic,hv_time' to cpu section of qemu invocation07:04
pedjavirt-manager does that automagically with Win guests07:04
frinnstmy biggest issues with win under qemu is the lack of working audio and shitty gfx performance07:56
frinnsthell, even running a linux vm without X results in shitty gfx. A scrolling terminal consumes 100% cpu07:59
frinnstnoticed that the other day07:59
*** chinarulezzz has joined #crux08:05
joacimi passed over the audio part of my graphics card too, so i get the displayport audio08:16
joacimpedja: works a lot better for eve online08:27
*** admincomplex has quit IRC08:43
joacimperformance with more demanding games is much lower still08:46
joacimin games where i would get 50 and up when running normally, i seems to get 20-30 FPS in qemu08:53
pedjamaybe AMD Windows drivers suck :)08:56
*** edvb has quit IRC08:58
pedjafrinnst, Win7 with ac97 and spice?09:09
pedjasome people say that qemu is still behind Virtualbox with Windows guests.09:10
pedja2d/3d performance09:11
frinnstwin7 x64 - no ac97 available on that arch09:11
frinnsthda is what's used and that seems to "kinda" work - i get audio noise in the speakers but its very distorted09:12
frinnstand yes, spice09:12
*** nwe has quit IRC09:30
*** nwe has joined #crux09:33
*** TheGallopingFox has left #crux ()09:38
*** cippp has quit IRC10:14
*** nwe has quit IRC10:17
*** nwe has joined #crux10:20
*** chinarulezzz has quit IRC10:28
pedjafrinnst, https://bugzilla.redhat.com/show_bug.cgi?id=1176761#c1610:46
pedjaand, when you are really bored, https://ladipro.wordpress.com/2017/02/24/running-hyperv-in-kvm-guest/10:49
pedjaserver 2k16 only10:49
*** heroux has quit IRC10:50
pedjathere are still a few bugs with AMD cpus, but it works10:55
pedjathe consensus seems to be that for the best Windows VM performance with Qemu you have to passthrough ALL THE THINGS11:00
*** xor29ah has quit IRC11:00
pedjaseparate hard disk, gfx card, preferably sound card, too11:01
pedjadual boot is much cheaper/faster/less fiddly11:02
pedjafor now, at least11:02
pedjatbh, I didn't really notice performance issues with Linux VMs, most of mine are servers which I ssh into11:04
*** xor29ah has joined #crux11:05
*** ileach has joined #crux11:15
*** heroux has joined #crux11:17
*** SiFuh has quit IRC11:19
*** cippp has joined #crux11:57
frinnstor just buy a another computer :)12:31
*** pejman has quit IRC12:48
*** pejman has joined #crux13:00
*** pejman has quit IRC13:00
*** pejman has joined #crux13:00
*** cippp has quit IRC13:11
*** cippp has joined #crux13:14
*** ileach has quit IRC13:20
*** pyuuun has quit IRC13:27
*** pejman has quit IRC13:28
*** pejman has joined #crux13:28
*** pejman has quit IRC13:28
*** pejman has joined #crux13:28
*** g0relike has joined #crux13:28
pedjapretty much :)13:29
pedja'I was going to go with Threadripper, but I couldn't justify needing all those cores.' no imagination whatsoever :)13:34
teK_make -j1613:34
fyredoes make not get any benefit using threads?13:50
*** pejman has quit IRC14:03
pedjaif I am reading this https://www.gnu.org/software/make/manual/html_node/Parallel.html correctly, not by default14:04
pedjaiow, you have to explicitly set it14:08
pedjaif there is an actual benefit is completely different question, and the answer depends on build system used, software itself14:09
jaegerfyre: I started doing some test recently along those lines just for grins: http://jaeger.morpheus.net/misc/llvm.png14:09
jaegerdepends a lot on the software, yeah14:09
fyrevery interesting14:10
pedjameson claims it uses all your cores/threads to the max14:10
fyredefinite decreasing benefits14:10
teK_why did you stop at a certain number of processes?14:10
teK_i.e. how does the i7 perform for >8?14:11
jaegerteK_: saw no point it using more threads than I had available, given the expected diminishing returns14:11
jaeger"more jobs than I had threads available" if you prefer14:12
teK_would be interesting how steep the costs for overloading threads are :]14:12
pedjaif would be interesting to see if there is any difference in build time when mesa switches to meson14:13
teK_btw what did draw the graph,  it doesnt look like gnuplot14:13
jaegerit's gnuplot14:16
*** g0relike has quit IRC14:16
teK_very nice14:16
jaegerJust had to give it some style. the default looked like this: http://jaeger.morpheus.net/misc/llvm-orig.png14:16
*** g0relike has joined #crux14:17
teK_before looking: it must have been looking very ugly if you saved the original14:17
teK_;-)14:17
teK_it did indeed14:17
jaegeryes :)14:17
teK_mind sharing the style / link? :]14:17
jaegerSure. I need to boot up the machine it was generated on anyway and add the i5 to the set14:18
pedja'Vega 56 delivers around 90% of Vega 64’s performance for 80% of the price'.14:18
teK_no hurry14:18
*** emmett1 has joined #crux14:21
*** pejman has joined #crux14:21
*** pejman has quit IRC14:21
*** pejman has joined #crux14:21
*** pejman has quit IRC14:28
*** SiFuh has joined #crux14:29
*** pejman has joined #crux14:29
*** pejman has quit IRC14:29
*** pejman has joined #crux14:29
*** pyuuun has joined #crux14:35
*** g0relike has quit IRC14:35
cruxbot[core.git/3.3]: start-stop-daemon: update version, signatures14:53
cruxbot[core.git/3.3]: start-stop-daemon.c update14:53
jaegerapparently I've lost the original command that plotted all three together but it won't be hard to recreate. The data is all there15:06
teK_not even in your shell history?15:06
joacimnowhere in your shells history?15:07
jaegerIt might be but it'll look like './plot' or something :P15:07
teK_hehe15:07
jaegerI still have the template I used to plot the first two, so I can recreate it from that. Just won't remember the exact color for the third data set15:07
jaegerwhich is no big deal15:07
teK_I have setopt histignorespace which will not put commands in my hist file if they start with a backspace15:08
cruxbot[opt.git/3.3]: gnupg: 2.1.21 -> 2.1.2315:08
jaegerI might make this stuff cleaner/prettier and upload it to github or something15:09
jaegeranyway, the template looked like this with 2 data sets: http://sprunge.us/gcKb15:11
*** g0relike has joined #crux15:34
*** pyuuun has quit IRC15:35
*** pyuuun has joined #crux15:46
*** g0relike has quit IRC15:46
*** jue has quit IRC16:07
*** jue_ has joined #crux16:08
*** onodera has joined #crux16:11
*** BitPuffin|osx has joined #crux16:27
pedjarejoice, jaeger, looks like docker has yet another dependency :)16:51
pedjadocker-cli16:52
pedjait almost seems like they are, not so gently, discouraging people not to build it from source themselves16:53
pedja'get it from the Docker store'. heh16:54
pedjathey are taking this modular thing pretty seriously, I'd say16:55
onoderaHas anyone here build 4.12 using BFQ as the default IO scheduler?17:04
*** pyuuun has quit IRC17:04
*** g0relike has joined #crux17:05
*** pyuuun has joined #crux17:06
*** g0relike has quit IRC17:06
pedjame17:06
pedjawell, I set it as default in rc.local17:08
onoderahmm when I used the patch I used to set the default in menu config, but now that it's in mainline it no longer in the "default scheduler" menu for some reason...17:08
pedjaif you use multiqueue, you can't set default scheduler that way17:09
pedjanot yet, anyway17:09
*** emmett1 has quit IRC17:10
onoderaso what is multu queue exactly? I've seen it a few times now?17:10
pedjaonodera, https://lwn.net/Articles/720675/17:12
pedjaSCSI_MQ_DEFAULT has to be enabled before BFQ and Kyber appear in IO schedulers part of kernel config17:16
pedjait's a bit confusing, CFQ is set as default, but on reboot mq_deadline was set for sda/sdb :)17:17
pedjathat's the default17:18
pedjait's a WIP, so :)17:19
onoderapedja: I enavled that setting but it still doesn't show up in the defau;t kernel list :/17:21
pedja'set default scheduler'17:21
pedja?17:21
onoderaye17:21
onoderabefore when I used bfq in patch from it would show up in there17:22
onoderanow with 4.12 it doesn't17:22
*** tilman has quit IRC17:23
pedjaone of the kernel devs removed setting default scheduler selection for blk-mq17:27
pedjaand blk-mq will be on by default in 4.13, iirc17:28
*** tilman has joined #crux17:28
pedjayou can use 'elevator=whatever' in grub. for now, that will be gone soon, too :)17:29
pedjaor an udev rule17:31
*** q6AA4FD has quit IRC17:32
*** q6AA4FD has joined #crux17:33
jaegerI should probably switch the docker port to the stable release rather than the edge release, maybe17:42
jaegerSince the edge releases are only supported for a month17:42
jaegerNot that they'd be supported on crux anyway, since we build from source17:42
*** pyuuun has quit IRC17:44
onoderapedja: can  you show me your rc.local please?17:45
onoderaor well, the relevant bfq line17:45
pedja"/bin/echo bfq > /sys/block/sda/queue/scheduler"17:46
pedjajaeger, but, but, what about new features? :)17:47
pedjadocker is a PITA17:48
*** onodera has quit IRC17:50
pedjaArch has a huuuuge pkgbuild thingie for it, all parts jammed into one17:50
pedjawhich sort of makes sense, since they all have to be updated at once, using specific commit17:52
pedjaI do something similar for vulkan-sdk17:52
*** onodera has joined #crux17:53
pedjait's ugly af, but it works17:53
*** g0relike has joined #crux17:53
pedjaonodera, that line works, but udev rule is much more elegant, imho17:56
*** JanC_ has joined #crux17:57
onoderapedja: how so, you mean newly mounted drives will also use bfq automatically? or is there more to it?18:05
pedjapretty much, yes. or if you have several ssds and spinning rust disks, automagically set scheduler for all of them at once18:07
pedjasomething like 'ACTION=="add|change", KERNEL=="sd[a-z]", ATTR{queue/rotational}=="1", ATTR{queue/scheduler}="bfq"'18:08
pedja(I stole that line from jaeger, I think :) )18:10
*** chinarulezzz has joined #crux18:11
*** pyuuun has joined #crux18:20
*** g0relike has quit IRC18:20
*** edvb has joined #crux18:26
*** groovy2shoes has quit IRC19:50
*** vitia69 has joined #crux20:07
*** vitia69 has quit IRC20:10
*** SiFuh has quit IRC20:42
*** SiFuh has joined #crux20:43
*** frinnst has quit IRC20:58
*** frinnst has joined #crux21:00
*** frinnst has quit IRC21:00
*** frinnst has joined #crux21:00
*** j_v has joined #crux21:11
*** john_cephalopoda has joined #crux21:24
*** SiFuh has quit IRC21:56
*** SiFuh has joined #crux21:58
*** s-mutin has quit IRC22:12
*** s-mutin has joined #crux22:13
*** TheCephalopod has joined #crux22:15
*** john_cephalopoda has quit IRC22:18
*** TheCephalopod has quit IRC22:39
*** s-mutin has quit IRC22:55
*** s-mutin has joined #crux22:55
*** chinarulezzz has quit IRC23:05
*** cippp has quit IRC23:07
*** cippp has joined #crux23:08
*** darfo has quit IRC23:51
*** ryu0 has quit IRC23:51
*** ][_R_][ has quit IRC23:51
*** duncaen has quit IRC23:51
*** heroux has quit IRC23:51
*** joacim has quit IRC23:51
*** cruxbot has quit IRC23:51
*** bedis has quit IRC23:51
*** SiFuh has quit IRC23:51
*** rwaweber has quit IRC23:51
*** rmull has joined #crux23:51
*** yui^^ has joined #crux23:51
*** SiFuh has joined #crux23:52
*** heroux has joined #crux23:55
*** joacim has joined #crux23:55
*** cruxbot has joined #crux23:55
*** bedis has joined #crux23:55
*** Introoter has joined #crux23:55
*** Introoter has quit IRC23:55
*** darfo has joined #crux23:56
*** ryu0 has joined #crux23:56
*** ][_R_][ has joined #crux23:56
*** duncaen has joined #crux23:56
*** jdolan has joined #crux23:56
*** Stealth has joined #crux23:56
*** rwaweber has joined #crux23:56
*** z3bra has joined #crux23:56
*** jaeger has joined #crux23:56
*** Stealth has quit IRC23:57
*** heroux has quit IRC23:57
*** joacim has quit IRC23:57
*** cruxbot has quit IRC23:57
*** bedis has quit IRC23:57
*** Introoter has joined #crux23:57
*** Introoter has quit IRC23:57
*** Introoter has joined #crux23:57
*** heroux has joined #crux23:59
*** joacim has joined #crux23:59
*** cruxbot has joined #crux23:59
*** bedis has joined #crux23:59

Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!