Werner changed the topic of #armbian to: armbian - Linux for ARM development boards | www.armbian.com | Github: github.com/armbian | Commits: #armbian-commits | Forums Feed: #armbian-rss | Type 'help' for help | This channel is logged -> irc.armbian.com
ChriChri_ has joined #armbian
ChriChri has quit [Ping timeout: 260 seconds]
ChriChri_ is now known as ChriChri
drobo_00 has quit [Quit: drobo_00]
sassinak-work has quit [Ping timeout: 240 seconds]
sassinak-work has joined #armbian
TRS-80 has joined #armbian
<TRS-80> There he is
<TRS-80> I knew I should come around this time :D
<Tonymac32> look what the cat dragged in
TheNumb has quit [Ping timeout: 265 seconds]
TheNumb has joined #armbian
<Tonymac32> TRS-80 cheers
<Tonymac32> lanefu my Pine H64-B has been running protein folding for several months with no downtime
<TRS-80> There he is! What's happening my negro
dddddd has quit [Ping timeout: 244 seconds]
<TRS-80> How are you getting on with OpenHAB?
<Tonymac32> Man I got so far, then got distracted. Story of my life
<lanefu> Tonymac32: sweet thats what i wanna hear... does it use a weird power connector or the opi kind?
<TRS-80> Did you really get far? Dude I am on like my third try, finally making good, solid and stable progress (I think?) lol
<Tonymac32> lanefu it uses the skinny 5V barrel jack
<Tonymac32> trs-80 I got to light flipping, some events, etc
<TRS-80> I have certain lights now coming on at sundown! Much excite!
<Tonymac32> nice
<Tonymac32> I need to pick a machine that isn't any good for anything else *OPi-3 1 GB COUGH*
<Tonymac32> and set it up for real
<Tonymac32> (Not a knock on the OPi-3, but mine is a 1 GB model, which is useless for a lot of things
<lanefu> Tonymac32: ha.. i have oneplus... i run pihole and wireguard on it
<lanefu> 1GB + java? :P
<TRS-80> yeah, Cubietruck with OpenHAB was a little sluggush at times, ODROID-XU4 now never breaks a sweat
<fizikz> Tonymac32: lanefu: odroid forum site admin thinks DRAM controller config could have issues on kernel 5 https://forum.odroid.com/viewtopic.php?f=184&t=38774
<lanefu> sorry i forgot not eeverything java is an enterprise spring ap
<lanefu> fizikz: good find.. yeeah i tested painline 5.7 and had same expeerience
<Tonymac32> lanefu headless with ZedRAM should be OK
<TRS-80> so I have Cubietruck running mosquitto and XMPP server and someother lightweight things now
<Tonymac32> fizikz that doesn't sound impossible
<lanefu> Tonymac32: would dram timings be device tree stuff?
<Tonymac32> the Cubietruck. Never has one. Was trying to get one of those Seeed iMX6 boards, but it disappeared in Cambodia according to the shipping paperwork
<Tonymac32> lanefu probably uboot stuff
<Tonymac32> a lot of these boards just trust that uboot configs the hardware
<TRS-80> Cubietruck is ancient history by now. You can get ODROID-XU4 actually cheaper anyway! I just had one for long time now.
<TRS-80> Fun fact: Igor trying to get/build Debian on a Cubietruck was what led to Armbian...
<Tonymac32> right. my home media/file server is an XU4
<TRS-80> could probably add OpenHAB to that no problem, unless you are like transcoding or something
<TRS-80> I was just putting finishing touches on custom MySensors / MQTT based thermostat / HVAC control! Much excite!
<Tonymac32> that does happen
<TRS-80> well you probably have multiple of them I would imagine, no? :)
<TRS-80> even I have 2... ;)
<Tonymac32> I may or may not have a stack of 5 MC1-solos....
<TRS-80> I do know that
<Tonymac32> and an XU4 running retroarch in my living room
* lanefu waves at Tonymac32
<TRS-80> Do you also stream? Or only video games?
<lanefu> still have 2 mc1-solos in plastic
<Tonymac32> I think the OPi-3 will be the one though, I honestly can't think of any other use I have for it since sunxi is not my wheelhouse
<TRS-80> bah, I keep forgetting I bought that damn HiFiBerry AMP2 for the RPi, guess it stays in living room
<fizikz> has anyone verified if the benchmark issues also exist on kernel 4.* lately?
<TRS-80> I mean, if you have it laying around... I suppose you are excused XD
<Tonymac32> fizikz no, I don't have any running that
<fizikz> me neither. but i'm thinking it would be good to check
<lanefu> going to roll to legacy kernel... looks like all armbians for xu4 use same uboot
<lanefu> Tonymac32: if i run kernel 4.9 on my N2 am I a quiter
<Tonymac32> ewwwwww
<Tonymac32> why
<lanefu> cuz 4k works
<Tonymac32> I mean if your allwinner stuff has 3.14 on it why not. :D
<lanefu> yah tritium h5 was driving my 4k last summer
<lanefu> on mainline
<Tonymac32> do you believe carburetors are superior to fuel injection by chance?
<lanefu> for throttle response
<Tonymac32> ... I assume a returned fuel system at all times
<Tonymac32> returnless is crap
<lanefu> ROFL
<Tonymac32> PWM fuel pump = sadface
<lanefu> ewww
<lanefu> even my carb has a return on the fuel pump
<lanefu> OMG i lost the needle to the carb on my powerwasher in the lawn
<lanefu> that was a fun search
<Tonymac32> hahahaha
<Tonymac32> so my wife's "Kohler Courage" pushmower wouldn't run with it's obviously not made-by-harbor-freight chinese engine
<Tonymac32> if it were a "Predator" branded Chongqing Rato or loncin it would be perfect
<Tonymac32> anyway, this thing is trash. I simply disassembled and reassembled it and it runs again, didn't find anything in the carb
<TRS-80> magic
<lanefu> lucky
<TRS-80> lol
<lanefu> i ended up getting a new carb, which has the wrong choke linkage
<Tonymac32> ugh
<lanefu> cuz mine is a Subaru/Robin engine
<lanefu> so parts are extra dumb
<Tonymac32> I know a Suburu SOUNDS like a lawn mower
<lanefu> hahaha yeah
<Tonymac32> that doesn't mean lawn mower engines are Subaru
<Tonymac32> :D
<lanefu> i had a 1974 VM karmann ghia with a 1776cc motor, then a 2110cc motor that i sometimes drove with open headers
<lanefu> it would at least someetimes sound like a harley
<Tonymac32> man, you really just go looking for pain and suffering I see
<Tonymac32> :P
<lanefu> yeah cuz you know what
<lanefu> the stock 1600cc with headers ran perfect
<lanefu> idled in 15 degree weather and had head
<lanefu> got like 30mpg
<lanefu> but nooo.. i wanted to go faste
<lanefu> r
<Tonymac32> hahaha
<Tonymac32> my brother got an Audi RS3
<Tonymac32> and immediately flashed it to take E85
<Tonymac32> so now fuel is an easter egg hunt and he has to monitor the quality of the fuel
<lanefu> yeah so i don't understand the E85 thing with drag motors
<Tonymac32> power
<Tonymac32> but it also drinks faster
<lanefu> i thought ethanol has less power
<Tonymac32> per volume yes
<Tonymac32> but you add lots
<lanefu> got it so it like atomizer and kabooms better
<lanefu> s/atomizer/atomises
<ArmbianHelper> lanefu meant to say: got it so it like atomises and kabooms better
<Tonymac32> in any case, to top it off it's a VW, and at $65k new it is about Buick levels of quality.
<Tonymac32> and not like high end buick quality
<Tonymac32> I mean like mid 90's buick Century
<lanefu> ROFL
<Werner> Good morning
<Tonymac32> It's everything I would expect of an economy luxury offering, other than the fact it is unquestionably fast
<Tonymac32> and not economy
<Tonymac32> Good morning
<lanefu> fizikz: testing Linux 4.14.186-odroidxu4 now
<lanefu> Werner: extra early for you
<Werner> Just arrived at my office
<fizikz> +1 i'm curious
<lanefu> Werner: what time do you go home from work
<Werner> lanefu: Somewhen between 2 and 3 pm cest. Sometimes earlier, sometimes later :P
<lanefu> 2 or 3PM is when people start to leave me alone and i can start actually doing work lol
<Tonymac32> I am legally required to work only 32 hours/week now, it is... interesting
<Tonymac32> since I was easily doing 50-60 before
<lanefu> Tonymac32: wow thats like type email walk away mid
<Tonymac32> close
<Werner> Tonymac32: Same thing, but both numbers are less atm
<Tonymac32> meetingsmeetingsmeetingsmeetingsmeetingsmeetingsmeetingsmeetingsmeetingsmeetingsmeetingsmeetingsmeetingsmeetingsmeetingsmeetingsmeetingsmeetingsmeetingsmeetings
<Tonymac32> lol
<Werner> Meeting...just a fancy word for what we call here "blabla" :P
<Tonymac32> after the necessary and completely mandatory 30 hours of meetings not much gets done ;-)
<Tonymac32> werner we have a much less appropriate term for it
<Tonymac32> :D
<Werner> Haha
<lanefu> hey do you all use custom feeds for the forums or anything
<lanefu> like tvboxes has ruined the forums for me
<Tonymac32> no, and yes
<lanefu> so i kinda dont look at the firehose anymore
<Tonymac32> I have this problem also
<Tonymac32> TVBian
<Tonymac32> :P
<Werner> I managed quite decently for myself to ignore it :D
<Tonymac32> but your avatar describes me looking at it
<Tonymac32> time to make Amlogic boards sing again, and also I have this ASUS board with literally the most massive power delivery system I've ever seen on an SBC
<lanefu> HHAHA @tony i could make some new membership groups and deny ourselves access to the tvboxes forum
<Tonymac32> rofl but sometimes I need to cheat off of Oleg
<Tonymac32> :)
<lanefu> yeah doubledged sword i guess
<lanefu> tell me about this asus board
<Tonymac32> RK3399 PRO
<lanefu> whats diff between pro and regular 3399
<Tonymac32> Has an NPU 4 GB RAM for SoC, 2 GB RAM for NPU
<lanefu> same core config, but NPU?
<Tonymac32> HDMI, DP over USB-C
<Tonymac32> wireless AC adapter
<Tonymac32> the input is 12V to a 5.5x2.5 barrel jack, so bigger center conductor than the XU4 for example
<Tonymac32> good for amps
<Tonymac32> and has a power input header
<Tonymac32> and the FETS used to protect against back-powering/reverse/etc are gigantic
<lanefu> sounds legit
<Tonymac32> jumpstart your small car massive
<lanefu> you're that excited about a mosfet that you shard the datasheet lol
<Tonymac32> OK, so everything the Tinker is terrible at, this thing just massively overcompensates for
<Tonymac32> I'm terrified to learn the $
<lanefu> wtf. does that really say 78 Amps
<Tonymac32> yes, but I think that is because they were taking some of our feedback seriously. look at the RDS_ON
<Tonymac32> so handling that kind of amperage doesn't allow for any losses in the device or it will catch fire
<Tonymac32> so the resistance at "on is reallllly low, meaning absolutely minimal voltage drop
<Tonymac32> Also, nice little heat sink/fan
<Tonymac32> seems effective
<Tonymac32> The key points are honestly the power input, the NPU, and the swappable wifi and slot for an LTE modem/etc on the bottom
<Tonymac32> other than that it's just another RK3399
<lanefu> good power goes a long way in my book
<Tonymac32> with that power system I'd feel comfortable actually rolling that into a customer-facing project.
<lanefu> i see why youre excitd now
<Tonymac32> I haven't tested yet, but the hype seems to claim you should be able to load the USB's to specification
<lanefu> DDR4 ram
<Tonymac32> unlike my N2 that can't handle an SSD and a wifi card without locking...
<lanefu> yeah n2 really seems to be a bust on that front
<lanefu> F@#$#@$
<lanefu> dude
<lanefu> dude
<Tonymac32> TBH this would be the first board I know of that could handle specification loads on USB. So I need to test
<lanefu> my active hdmi->display port adapter on my n2
<lanefu> ....powered by n2
<Tonymac32> hehehehe
<lanefu> i wont get that lucky of course
<Tonymac32> I was a pissed about that as I was when I realized my OPi3 only had 1 GB RAM
<lanefu> ha
<lanefu> dude my opi3 has been a champ
<Tonymac32> If a board is lacking an HDMI, 1 GB ok
<Tonymac32> but that thing is intended to be full-feature
<lanefu> yeah.. 1gb opi3 is basically the opi one plus where they actually gave you usb3
<Tonymac32> but guess how it's going to get used? headless
<lanefu> yeah headless is when the fun begins
<lanefu> was talkign to igor we need a desktop czar
<Tonymac32> speaking of, I think my MC1's will be a build farm for my embedded work
<lanefu> which i guess should have been oleg
* Tonymac32 cries in Cyrillic
<lanefu> yeah hopefully decent esp32 toolchain for arm now
<lanefu> i had to run esphome on x86 cuz esp32 builds
<Tonymac32> Yeah, that work is a bit separate because I can't export that stuff myself (yet)
<lanefu> gotcha... well what's tht chip agian on your little zero formfactor board
<Tonymac32> that's why I went with CrowdSupply, who is still just ridiculous and not following through on their side
<Tonymac32> the first one is a 328PB, so the newer brother of the UNO chip
<lanefu> yeah i guess they just are used to selling star trek keycaps
<Tonymac32> a couple extra bells and whistles, more IO
<Tonymac32> the other one is a CH559, an 8051-based USB MCU
<Tonymac32> that one uses SDCC
<lanefu> Software Defined Credit Cards
<Tonymac32> I have it working with PlatformIO, but they don't seem to care about my pull request :P
<lanefu> man thats not cool
<lanefu> i use platofrm IO for building tasmota firmware
<lanefu> just nice I can use vim
<lanefu> lol
<lanefu> emory performance (big.LITTLE cores measured individually):
<lanefu> memcpy: 379.8 MB/s
<lanefu> memset: 801.4 MB/s (0.2%)
<lanefu> memcpy: 2423.4 MB/s
<lanefu> memset: 4908.7 MB/s (0.9%)y
<lanefu> fizikz: http://ix.io/2qxh
<Tonymac32> I'll double-check that I didn't miss an etiquette point about how to open a PR
<Tonymac32> but
<lanefu> grrr whitequark is down
<lanefu> i need my benchmakr urls from yesterday lol
<Tonymac32> lol
<Tonymac32> I have committed to memory, memcpy was at 80 MB/s
<Tonymac32> +/- 4 MB/s
<fizikz> lanefu: nice confirmation. 380MB/s memcpy and better latencies on 4.14.186-odroidxu4 (mcsolo-1)
<fizikz> and not much throttling :)
<fizikz> do you want to post the result on the odroid forum thread for future reference, or would you like me to do it?
<Tonymac32> I am curious why the DRAM controller would have different settings to/from the big or LITTLE ones
<fizikz> Tonymac32: i wondered about that too, but i'm not knowledgeable on these details
<Tonymac32> me either really
<Tonymac32> The last time I really understood RAM you could hand solder in. :D
<fizikz> at least there's been progress in that area :D
<ArmbianHelper> AR-337 [Bug] "Odroid XU4 Memcopy Slow on all Kernel 5.x 80MB/sec instead of 370+MB/sec" reported by Lane Jennison at 2020-06-30. Status: To Do
<Tonymac32> https://github.com/platformio/platformio-core/blob/develop/CONTRIBUTING.md le sigh. I don't know if this applies to the other cores, but I'll look a bit deeper
<fizikz> awesome. do you think the latencies are also a symptom/hint?
<Tonymac32> if it's the dram controller, the latencies are probably the cause
<fizikz> so that should have an impact on overall performance even on big cores? even though the throughput on big cores seems unaffected?
<Tonymac32> well I don't know how the RAM is coupled to the controller to the cores, so unknown
<lanefu> with solder
<fizikz> lol
sunshavi has quit [Read error: Connection reset by peer]
sunshavi has joined #armbian
<lanefu> so some dynamic memory timing patch was added around kernel 5.3
<lanefu> we're using it
<lanefu> config/kernel/linux-odroidxu4-current.config:CONFIG_EXYNOS5422_DMC=y
<lanefu> config/kernel/linux-odroidxu4-dev.config:CONFIG_EXYNOS5422_DMC=y
<lanefu> building a kernel with that disabled
<lanefu> fizikz: know how to just to the memcopy test by any chance?
<lanefu> we cheeckout from odroid's branch/ its possible they addeed patch sooner
<Werner> maybe
<Tonymac32> There was a lot of shuffling around because of that dmc before it got put into the mainline
<Tonymac32> so could be
<lanefu> wee dont seem to havee any DMC patchees
<lanefu> tinymembench is holding me hostage
<lanefu> HA!
<lanefu> Memory performance (big.LITTLE cores measured individually):
<lanefu> memcpy: 304.6 MB/s
<lanefu> memset: 799.8 MB/s (0.1%)
<lanefu> memcpy: 2280.3 MB/s
<lanefu> memset: 4834.7 MB/s (0.6%)
<chewitt> I spotted someone mentioning that audio was broken in 5.7.x for Amlogic the other day (or night)
<chewitt> 5.7.6 has an accidental backport from 5.8 included that needs to be reverted
<chewitt> commit is 1bb707fbfd5c246028d76b8f11a19dfd118d6306
<chewitt> 5.7.5 and earlier are okay
<chewitt> I believe a fix will come, but until then the revert is fine
<lanefu> time for me to crash. hopefully thats an easy cherry-pick to pull out
<fizikz> lanefu: sorry, no i don't know if memcpy can be run on its own. but sounds like you made an exciting breakthrough!
sassinak-work has quit [Ping timeout: 240 seconds]
sassinak-work has joined #armbian
indy has joined #armbian
RzR has joined #armbian
psydruid has quit [Quit: killed]
JuniorJPDJ has quit [Quit: killed]
agates[m] has quit [Quit: killed]
hyperreal has quit [Quit: killed]
solderfumes[m] has quit [Quit: killed]
nekomancer[m] has quit [Quit: killed]
TRS-80 has quit [Quit: WeeChat 2.8]
hyperreal has joined #armbian
solderfumes[m] has joined #armbian
agates[m] has joined #armbian
JuniorJPDJ has joined #armbian
phantomse[m] has joined #armbian
psydruid has joined #armbian
nekomancer[m] has joined #armbian
Nakaori has quit [Ping timeout: 256 seconds]
Dr{Who} has quit [Remote host closed the connection]
Dr{Who} has joined #armbian
phantomse[m] has quit [Quit: Idle for 30+ days]
Nakaori has joined #armbian
xecutertool has quit [Remote host closed the connection]
xecuter has joined #armbian
c0rnelius has quit [Quit: WeeChat 2.3]
<Werner> Nice catch indeed if it is true.
xecuter has quit [Remote host closed the connection]
xecuter has joined #armbian
MrFixIt has quit [Ping timeout: 240 seconds]
MrFixIt has joined #armbian
<lanefu> fizikz: you wanna test kernel for stability?
awf_sbrt has quit [Read error: Connection reset by peer]
<lanefu> IgorPec: Werner: Is there any sort of syntax in Jira for annotating something as a "known issue"
<lanefu> that could go into release notes easily?
<Werner> Not sure what you mean. The Jira IRC bot?
<IgorPec> not that i am aware
<lanefu> nah, Jira Jira
<lanefu> k
<IgorPec> but how do you mean to tag this or what?
<IgorPec> to display known issues here?
<IgorPec> on IRC?
<lanefu> not IRC.. mostly just thinking about how to easily log something so that it shows up as a "know issue" with release notes
<lanefu> s/know/known
<ArmbianHelper> lanefu meant to say: not IRC.. mostly just thinking about how to easily log something so that it shows up as a "known issue" with release notes
<IgorPec> ahaa, you mean like a new group: "Story" "Fixed Bugs" ... "Known issues"?
<lanefu> actually yeah
<lanefu> cuz we can liknk a story, or a bog, to a known issue too
<lanefu> s/bog/bug
<ArmbianHelper> lanefu meant to say: cuz we can liknk a story, or a bug, to a known issue too
<IgorPec> probably by adding a new type ... let me test
<IgorPec> it is possible to add "New issue type" ... but am not sure if this is the best way
<IgorPec> perhaps displaying opened bugs from the backlog under is better?
<IgorPec> if possible to automatise
archetech has joined #armbian
<lanefu> IgorPec: yeah good point... maybe managing the bugs better makes more sense
<IgorPec> because adding more types will make more confusion, i guess
<IgorPec> in theory all bugs that are recorded are known
<Tonymac32> thanks chewitt I'll take a look later today (day job ATM)
<Tonymac32> lanefu like I said there was a lot of back and forth about that DMC before it was accepted, it wouldn't be that surprising to find something is messed up still. Good catch
<lanefu> Tonymac32: thanks.. classic case of open source group effort... since it took fizikz and odroid's response for clues. and you emphasizing how terrible 80MB is :P
<lanefu> IgorPec: So here's kind of an example of what i'm trying to figure out to track... We found a bug, we have a fix, but it's probably considered a workaround... so hwo to keep track and document that its a workaround https://armbian.atlassian.net/browse/AR-337
<ArmbianHelper> AR-337 [Bug] "Odroid XU4 Memcopy Slow on all Kernel 5.x 80MB/sec instead of 370+MB/sec" reported by Lane Jennison at 2020-06-30. Status: To Do
<IgorPec> lanefu: either with a label or component
<lanefu> k... label probably makes most sense
<IgorPec> yes. we don't use this much anyway
<IgorPec> done
Ntemis has joined #armbian
indy has quit [Ping timeout: 265 seconds]
<lanefu> okay now that I made a difference.. back to messing with CassandraDB :(
<Tonymac32> haha
drobo_00 has joined #armbian
<lanefu> i like how my code change is literally 2 characters... (actually the same character twice).. and then I write this like opus of ticket info
indy has joined #armbian
<fromport> lanefu: i would like to test your kernel on my mc1/hc2/xu4 !
<IgorPec> fromport ... its getting to beta.armbian.com in 1h
<IgorPec> or less, its building right now
<fromport> nice!
<fromport> are you building that native on arm or are you using cross compiler ? (just curious)
<IgorPec> cross
<Tonymac32> Igor this may deal with the instability issues as well under heavy load
<IgorPec> possible, i will also do some more testing, have xu4 on the desk
<fromport> sbc-bench never finishes on my mc1 (no fan) and reboots
<lanefu> fromport: interesting.. i've been doing my testing on 2 MC1's
<Tonymac32> BOINC always crashes at the end of task, machine hangs for me. I couldn't find any warning for it so far, it just does
<lanefu> I have mixed feelings aboutu BIG.Little I know Jon Masters doesn't have much confidence in anybody implementing it properly
<lanefu> Tonymac32: man i wonder what would happen if you just shutoff the little cores when you ran boinc
<Tonymac32> I don't think it's that simple
<Tonymac32> unfortunately
<lanefu> it is if you don't understand how things work... like me :P
<IgorPec> disabling little cores might actually be a solution ... if we could do it
<IgorPec> easily
<Tonymac32> it can be done via the device tree
<Tonymac32> as far as I know
<IgorPec> probably ... workaround as a last restort
<lanefu> for ID in {0..3};do echo "0" > /sys/devices/system/cpu/cpu${ID}/online;done
<IgorPec> does it work?
<lanefu> yeah man
<lanefu> and echo 1 to turn them back on
<ArmbianHelper> great, adjective, synonym for Armbian
<IgorPec> .great
<lanefu> htop only shows 4 cpus now
<lanefu> lol
<lanefu> thee little cpus are first, right?
cstroie has joined #armbian
<lanefu> I don't think i'd wanna ship with cpus disabled.. but maybe as a tweak in armbian-config
drobo_01 has joined #armbian
<lanefu> de-littler
<lanefu> lol
drobo_00 has quit [Ping timeout: 258 seconds]
drobo_01 is now known as drobo_00
<lanefu> or could just put the trick in the docs
<lanefu> imma stop talking now
dddddd has joined #armbian
<ArmbianTwitter> @WanganTAMan (BUIZ♢): How do you even make the brother printer work in an armbian CUPS server- (9s ago)
<Tonymac32> honestly I think disabling the DMC is the way forward, it obviously defaults to high-speed, unlike the RK3328 stuff that defaults to Extra Lame Mode (TM)
<lanefu> ^^ Throw in trash buy HP
<IgorPec> haha
<Tonymac32> I have a brother laser printer, it goes into deep sleep and never wakes up
<IgorPec> i have HP and it works ;)
<lanefu> i have a HP Laser from 2003 and it just works
<Tonymac32> great feature for WiFi printing...
<Tonymac32> unfortunately you're out of touch with modern day
<Tonymac32> :P
<lanefu> hahaha
<lanefu> i like my bubble
* fromport is very happy with it's brother printer. PITA to get working often, but once it works, it is great. great costs/performance
<IgorPec> fromport: apt update and upgrade on beta repository will give you lates kernels
<fromport> IgorPec: thanks!
<lanefu> IgorPec: aka switch to "nightly" from "stable" in armbian-config?
archetech has quit [Quit: Leaving]
<IgorPec> yes
<IgorPec> now this scripts still runs on my command, but it could also run in a cron
<IgorPec> since it only goes to rebuild stuff if changed
archetech has joined #armbian
<fromport> Unpacking linux-image-current-odroidxu4 (20.08.0-trunk.28) over (20.08.0-trunk.27) ...
<IgorPec> yes
<lanefu> IgorPec: after makng changes to shellcheck last night.. it gave me some more thoughts about testing
<IgorPec> like?
<lanefu> well kind of tying into what you started with https://github.com/armbian/autotests/blob/master/README.md#to-dos
<lanefu> ideally we should collect all our test benchmarks so we can trend over time
<lanefu> and suggested feedback.. use json format rather than XML for data structure
<IgorPec> yes, this is planned. perhaps currently not matured enoughg
<IgorPec> and that new hardware toy for switching would be useful here
<lanefu> k i might play with it a bit.. should be pretty easy to post json records into something
<lanefu> yeah once that tester hardware is here, the metrics stuff will rock
<IgorPec> try to make a format that can be easly imported to wordpress
<lanefu> okay will do
<IgorPec> that this can swallow https://wordpress.org/plugins/wp-all-import/
<lanefu> okay yeah should be cake
<lanefu> gonna see if I can get one of our data engineers to help me a bit
<IgorPec> great. i was not doing much with xml creation
xecuter has quit [Remote host closed the connection]
xecuter has joined #armbian
sunshavi has quit [Ping timeout: 256 seconds]
cstroie has quit [Ping timeout: 246 seconds]
Hokedli has joined #armbian
Hokedli has quit [Quit: Konversation terminated!]
Hokedli has joined #armbian
<fizikz> lanefu: fromport: i don't mind quick tests, but my hc2 is my backup server so i don't want to take extended/big risks on it. thanks to fromport for stepping up with an odroid farm!
<fizikz> is it simple to downgrade from nightly back to stable later?
<lanefu> fizikz: yep you can do it all from armbian-config
<martinayotte> IgorPec: I think I've screwed up something in git : I'm seeing https://github.com/armbian/build/commit/f863010e75b15e688b0d7ebefe95fb7b4f846859#diff-e5c5e6dae38b284e8cf0bd1fb0efac03 but this was because I've submit a new path without doing "git pull" first.
<fromport> fizikz: am running latest kernel now on both mc1 & hc2. sbc-bench takes over an hour, will let you know ;-)
<fromport> hc2 just finished: http://ix.io/2qAk
<fizikz> lanefu: and there shouldn't be issues from downgrading back to stable after nightly? the beta packages will get removed cleanly?
<fizikz> fromport: exciting!
<fromport> standard memcpy : 327.8 MB/s <- much better ! thanks all!
<fromport> vs standard memcpy : 85.5 MB/s on my "old" kernel: http://ix.io/2qA6
<fizikz> fromport: and i see the memcpy result is fixed while you're still on the hc1 optimized board config and the resulting slower cpu frequencies
<martinayotte> IgorPec: Ok ! It seems only "history pollution", nothing changed in files histories ...
<fromport> fizikz: reconfigured and rerunning
<fizikz> nice. also note the latency results are fixed too
<fromport> my mc1 bench is still running, max 1.9Ghz as well, will change/retest when it is done as well. thanks for your eye for details ;-)
<fizikz> :)
<lanefu> fizikz: yeah you just switch back to stable from armbian-config
<lanefu> that's awesome news y'all
<fromport> thanks lanefu for figuring it out !
<fromport> hope this gets fixed "upstream"
<lanefu> fromport: yeah the real fix is for them to figure out the timing stuff.. fortunately this fix is just toggling a kernel feature so nothing to upstream
<fizikz> lanefu: is it armbian-config -> System -> Nightly ? switching back would have a Stable option? would the installed nightly packages get downgraded automatically?
<lanefu> s/this fix/this workaround/
<ArmbianHelper> lanefu meant to say: fromport: yeah the real fix is for them to figure out the timing stuff.. fortunately this workaround is just toggling a kernel feature so nothing to upstream
<lanefu> fizikz: correct
<lanefu> Werner: I f-ing love that sed thing with armbianhelper
<fromport> lanefu: i meant to say "i hope the upstream figures out why the patch is working on 4.x but not on 5.x kernels and i hope they fix it asap " ;-)
<lanefu> oh that patch doesnt exist in 4.x
<fromport> hmmm..
<fromport> did you notify the author directly or just the bug report ?
<fromport> afk, emergency at work. bbl
archetech has quit [Quit: Konversation terminated!]
<lanefu> fromport: I just made the armbian bug report... i also posted updates on the odroid forum thread fizikz started... https://forum.odroid.com/viewtopic.php?f=184&t=38774 hoping hardkernel will continue to take interest... i'mm not really up for eaching out to maintainer since i dont understand tehmemory timing stuff
<IgorPec> martinayotte: great its nothing
<martinayotte> Right ! It was happening to me in the past, but it is always scaring when I do such mistake.
<IgorPec> git is a powerfull tool and it takes time to master.
<IgorPec> i also sometimes just stare :)
<IgorPec> or google
<Werner> lanefu, It's a toy :D
drobo_00 has quit [Quit: drobo_00]
sunshavi has joined #armbian
Hokedli has quit [Remote host closed the connection]
drobo_00 has joined #armbian
sunshavi has quit [Ping timeout: 244 seconds]
<fizikz> lanefu: memory (318 MB/s) and latency benchmark results confirmed fixed with kernel 5.4.49-odroidxu4 on my HC2 http://ix.io/2qAV
<lanefu> fizikz: that's great news... thanks for speaking up and helping me understand the problem
<fizikz> now for the real test: doing borg check. if that finishes in 2-3 hours instead of 5 hours, this will be a happy day
<lanefu> Ohh yeah the grand finale
<ArmbianTwitter> @IpponUSA (Ippon Technologies USA): Mid-journey to the #cloud and have compute sitting idle in your #datacenter? See how @lanefu modernized a project’s testing to the cloud with #GitHubActions, while still using #onprem #compute to save on costs! https://t.co/e5BhkZKlXu #Armbian #Linux #GitHub #CICD #DevOps (7s ago)
<fizikz> thanks for actually working on this problem. it's been around for a few months and there was little to no response on it. admittedly a tough problem where the first suspect is borg or some other software that might have changed
<lanefu> :) :) woohoo my blog posted
sunshavi has joined #armbian
<ArmbianTwitter> @lanefu (Lane Jennison): I'm particularly proud of this blog as I was able to practice my #DevOps DevOps craft while benefiting @Armbian, a Linux project I'm very passionate about. https://t.co/VuMRLVzjh2 (7s ago)
Ntemis has quit [Read error: Connection reset by peer]
<steev> hmm
<steev> 5.7 doesn't want to boot on the neo4
<steev> (or it's booted but video and networking don't work properly)
<steev> also, this is a thing for 5.7 to keep in mind for users with external wifi devices - https://lkml.org/lkml/2020/6/22/633 (ath9k is broken)
xec has joined #armbian
xecuter has quit [Ping timeout: 244 seconds]
sassinak-work has quit [Ping timeout: 246 seconds]
sassinak-work has joined #armbian
<fizikz> lanefu: preliminary result looking good. borg check on one of the repos finished in under half the previous recent completion times. seems like this memory issue and fix/workaround are relevant to my borg problem
<lanefu> fizikz: saweeet. Nice to have really tangible improvements
<ArmbianTwitter> @NicoD99364191 (NicoD): How to install #Panfrost on #Armbian #Focal https://t.co/h4HpFnq6JC (24s ago)
<ArmbianTwitter> @armbian (armbian): RT @NicoD99364191: How to install #Panfrost on #Armbian #Focal https://t.co/h4HpFnq6JC (9s ago)
drobo_00 has quit [Quit: drobo_00]
xecuter has joined #armbian
xec has quit [Remote host closed the connection]
drobo_00 has joined #armbian
drobo_00 has quit [Quit: drobo_00]
c0rnelius has joined #armbian
<fizikz> lanefu: borg check completed in 2.75 hours, as in omv4 / kernel 4.x vs the lately 7-8 hours in omv5 / kernel 5.x. oh the joy...
<lanefu> Woot!
<lanefu> I hope the odroid folks followup on your forum thread
<fizikz> one can hope
<fizikz> any idea when this workaround will be pushed to stable?
<lanefu> Not sure. I'll cherry-pick a hotfix and we can merge when appropriate
sunshavi has quit [Read error: Connection reset by peer]
<fizikz> ok. in the meantime i'll just avoid installing updates from nightly on my backup server
sunshavi has joined #armbian
<fizikz> now for a minor quibble, any idea why htop is showing LITTLE for all 8 cores?
<lanefu> Try uninstalling, apt cache clear, then reinstall
<lanefu> Maybe your large cores never fully matured and they're still small?