IRC log of #cubox of Sat 21 Dec 2013. All times are in CET < Back to index

00:23 jnettlet hste, figures. I just downloaded the previous latest release before I walked out of the house.
00:33 dbsx jnettlet: can you recommend a uboot binary? my i4pro needs to be powered on dozens of times before I get a kernel loaded (a random number of attempts). Using SR 2013.10-rc4-g22f867a-dirty
00:46 jnettlet dbsx, that is the one that shipped with it, or did you download it from rabeeh's tarball? Right now each chip revision needs its own u-boot so you can't mix and match.
00:46 jnettlet I am just getting ready to crash for tonight but would be willing to build you a debug u-boot tomorrow and we can do some testing.
00:47 jnettle 00:47 * jnettlet afk
01:13 _rmk_ ok, assuming I don't find anything further in need of fixing... tomorrow I should be able to get back to cubox/hb hacking
01:27 dbsx jnettlet: thanks. I used the i4pro from rabeehs images at getting started
01:28 dbsx http://download.solid-run.com/pub/solidrun/cubox-i/GeexBox-xbmc/geexbox-devel-20131212-r16611.cuboxi/geexbox-devel-20131212-r16611.cuboxi-patched-cubox-i4pro.img.zip
12:15 hste rabeeh: ping
12:25 jnettlet hste, so it turns out the IMX port of XBMC was broken for use with X11. I fixed that this morning and am rebuilding now.
12:25 jnettlet although now I want to test the SDL2 patch that is floating around for it.
12:26 hste jnettlet: gone some hours on making xbmc in ubuntu for you :)
12:27 jnettlet hste, is that with framebuffer support or X?
12:28 jnettlet dbsx, I am around the computer for 2-3 hours if you want to beat on u-boot a bit
12:29 hste I dont remember the sdl2 patch
12:37 Matoking Are there any people here who have received their Cuboxes yet?
12:39 jnettlet Matoking, do you need to ask a specific question that only a Cubox owner could answer?
12:40 Matoking Yeah, I'd like to know what distros and applications people have gotten working
12:40 Matoking And uBoot, which I presume works with Cubox-i just fine?
12:40 _rmk_ you'll find documentation on the wiki
12:40 jnettlet I think the wiki and forums are the best place to get that information.
12:41 jnettle 12:41 * jnettlet has never heard of a product that ships without the ability to boot.
12:41 jnettlet _rmk_, glad to see you are back connected to the internets today. Speedtest running any better for you?
12:42 Matoking Oh, there are instructions for compiling it
12:42 jnettlet for those that prefer that method, sure.
12:43 hste jnettlet_ a person like u? :)
12:43 jnettle 12:43 * jnettlet has been known to compile a u-boot or two
12:43 Matoking I've compiled plenty of applications myself
12:43 Matoking Linux is actually really easy in that regard
12:44 jnettlet Matoking, no offense, we are all just a little giddy about the holidays
12:44 jnettlet Decking our halls with lots of ARM boards
12:44 _rmk_ jnettlet: no, it's utter crap
12:45 jnettlet _rmk_, what DSL hardware did they stick you with?
12:46 Matoking @jnettlet Are you an employee of Solid Run or am I misinterpreting you somehow? :P
12:46 _rmk_ no idea, it's their own VDSL modem
12:46 hste _rmk_: what isp do u have?
12:46 _rmk_ enta.net (we're a partner, so the fttc installation was free)
12:47 _rmk_ maktoking: there's only one person (afaik) here who's a solidrun employee and that's Rabeeh
12:47 jnettlet they really push that. Top reasons to partner with enta.net
12:47 _rmk_ the rest of us are volunteers
12:47 Matoking Ok
12:48 jnettle 12:48 * jnettlet prefers enthusiasts
12:48 Coolgeek early adopters maybe ?
12:48 Matoking Probably
12:49 jnettlet masochists
12:49 hste Matoking: do u got a box?
12:49 Matoking If I had known i2Ultra would have taken longer to ship I would've gone with i4Pro instead
12:49 Matoking @Matoking No, I don't have one yet
12:49 jnettlet Matoking, you can probably change your order. email the customer support address on the website.
12:50 Matoking @jnettlet I guess that would be worth a try :)
12:50 jnettlet Matoking, Solidrun is very accomodating, I am sure you can work something reasonable out.
13:14 Matoking @jnettlet Sent them an email message
13:14 jnettlet Matoking, good luck
13:14 Matoking Yeah, I was thinking of upgrading my order anyway
13:15 Matoking I recently upgraded my VPS from 1 GB of RAM to 2 GB and seeing the change in performance was enlightening to say the least :P
13:17 dbsx Ok. Here
13:18 dbsx Does anyone have console mode uboot working?
13:19 dbsx on the i4pro
13:19 jnettle 13:19 * jnettlet thinks most do. Although you may be the only one with an i4pro.
13:20 jnettlet the hardware works pretty much the same, just different ddr timings.
13:20 jnettlet the register changes are all handled generically
13:20 dbsx I get a serial console to appear for a little while and then everything stops. 1 out of 3 power ons get me to uboot
13:20 jnettlet meaning all the other MX6 boards use the same register changes, so they aren't Cubox-i specific
13:21 jnettlet dbsx, okay. Do you want a u-boot build with debug output enabled?
13:22 dbsx yes please. It is late my time. So I might not get it done until the morning.
13:22 dbsx I will see how I hold up
13:27 dbsx I could (maybe) build uboot. Can I use native cubox (not i) to build uboot as armv7 code?
13:27 jnettlet dbsx, actually that will need to hold a bit. I just realized I am missing a file from rabeeeh.
13:28 dbsx jn, thanks I will get some sleep then
13:28 jnettlet rabeeh, when you get a chance I am missing the include/configs/mx6_cubox-i.h machine file from your updates.
13:28 jnettlet dbsx, will do.
13:29 dbsx So is armv5t or armv7l OK for uboot on the imx6
13:29 dbsx ?
13:31 dbsx or do I really need to get a build env?
13:33 sickness what's the best and more stable java to use on cubox? i2p keeps exiting after about 24 hours with openjdk 7 :/
13:33 sickness JVM received a signal UNKNOWN (6).
13:37 jnettlet sickness, oracle has released an official "java" armhfp set of binaries
13:38 sickness so you suggest to try those? you mean java8 for arm?
13:39 jnettlet sickness, well that is really your only other option :-)
13:39 sickness yeah, I wouldn't like to go back to java 6... but I could try
13:40 jnettlet of course being a binary build and a US company I can't guarantee there aren't backdoors built into it for the NSA.
13:41 jnettlet just saying because you are running i2p
13:41 sickness yeah, in fact that was my main concern, it's not even opensourced :/
13:44 jnettlet sickness, are you dumping a core file. signal 6 is a SIGABRT. if you get a backtrace you can submit a bug.
14:25 hste jnettlet: how is it going with xbmc?
14:26 jnettlet hste, it built and ran into a bug with X11 and GLES which was why it was disabled by default. I have patched that and am linking now.
14:27 hste jnettlet: I guess it will not be far from memory limits on c1 running it in ubuntu and X
14:28 jnettlet hste, well we will see. I actually have zswap backported and enabled on my C1. Trying to see how much that helps.
14:45 dv so jnettlet, hows your 3.10 kernel doing ?
14:45 dv you mentioned something about VPU tests
14:47 jnettlet dv, I got the various freescale vpu libraries compiled and sorted yesterday. Been fighting with xbmc under X11 since.
14:47 dv jnettlet: imx-vpu, imx-lib, libfslvpuwrap you mean?
14:47 jnettlet I thought about just testing with gstreamer, but then got stubborn
14:47 jnettlet yep
14:47 dv yeah thats a bit messy
14:48 hste I got gplay and te yocto app video to work. its using the 0.10 version
14:48 d 14:48 * dv can now continue to work on a revamped eglvivsink
14:48 dv finally
14:49 hste :)
14:49 jnettlet xbmc has started under X
14:49 jnettlet hurrah
14:52 jnettlet not too bad cpu usage looks to be between 25 and 50% running the gui
14:52 hste dirtyregions?
14:52 dv you think I can include your kernel in the OE recipes already?
14:53 dv now that the name is chosen, I think I can send otavio patches
14:54 jnettlet dv, I need to sort out audio still. I think so. I have the zswap option enabled, but it is off by default without a kernel option.
14:54 jnettlet hste, how do I check dirtyregions?
14:54 hste is in Advancedsetting.xml if I remember
14:55 jnettlet okay so it does have the cpu clocked all the way up to 1ghz, and is rendering 1080p at between 13 and 20fps
14:56 hste jnettlet: what bootcmd options are u using?
14:56 jnettlet for xbmc?
14:56 hste yes
14:56 jnettlet none
14:56 jnettlet just fired it up
14:57 jnettlet just raw power baby!
14:58 _rmk_ jnettlet: have you done any iperf measurements on the cubox and hb ?
14:58 jnettlet _rmk_, not yet no. It is on my list for the weekend.
14:59 jnettlet I am running all sorts of benchmarks over the weekend for another project. So far nothing extraordinary has popped up.
14:59 hste What does the No such IOCTL, cmd is 22032 message mean?
14:59 _rmk_ jnettlet: this is between my laptop and the hb:
14:59 _rmk_ [ 4] local 192.168.1.181 port 5001 connected with 192.168.1.11 port 39877
14:59 _rmk_ [ 4] 0.0-10.0 sec 543 MBytes 455 Mbits/sec
14:59 _rmk_ [ 4] local 192.168.1.181 port 48258 connected with 192.168.1.11 port 5001
14:59 _rmk_ [ 4] 0.0-10.0 sec 158 MBytes 132 Mbits/sec
14:59 _rmk_ .181 is hb, .11 is the laptop
15:00 _rmk_ and... port 5001 is the receiver of the traffic
15:00 dv _rmk_: reminds me of that article about imx6 ethernet performance
15:00 _rmk_ so we're pretty slow at sending :(
15:00 dv http://boundarydevices.com/i-mx6-ethernet/ <- this one
15:00 jnettlet transmit runs on a separate clock right?
15:01 _rmk_ yea but it can't be that far out :)
15:01 _rmk_ I suspect it's the dma buffer flushing
15:01 dv they reach considerably higher performance .. and less difference between sending and receiving, so perhaps it can be tweaked further
15:02 hste what about irq balancing?
15:02 _rmk_ my hb is a solo device
15:02 hste :)
15:02 _rmk_ and if you want irq balancing on ARM, run the irqbalance daemon :)
15:03 jnettlet that is pretty much the case for any platform.
15:04 _rmk_ yep - I refuse to have code in the kernel to do that, because getting the algorithm right is not easy (I've tried various solutions...) and since it's already been solved via irqbalanced, there's no point wasting effort on doing it in kernel space
15:04 jnettlet I did some work for a company not to be named, and they couldn't replicate my metrics on their "big iron". Turns out their Chinese developers removed irqbalanced from my image as it was deemed "unneccessary"
15:05 jnettlet doh
15:05 hste jnettlet: dirtyregions is in .xbmc/userdata/advancedsettings.xml
15:06 jnettlet okay whatever xbmc is trying to do to playback video is not working. I get the control overlay, but no video and my mouse cursor gets duplicated on the screen as I drag it around
15:06 hste jnettlet: but u got a nice gui at least :)
15:07 _rmk_ there's two conflicts: 1. you want to keep any particular irq on the same CPU as much as possible and not bounce it around since it will benefit from the local CPUs cache. 2. you want to distribute the IRQs across the CPUs particularly busy ones so that they're not all on the same CPU.
15:07 jnettlet hste_ and more importantly a GLES accelerated gui running 1080p on the C1 using < 50% cpu
15:08 jnettlet oh in this branch there is a patch that is limiting fps. I don't think I want that yet.
15:09 jnettlet go big or go home!
15:09 hste jnettlet: do u use fsl 3.5.0.7 or 3.10.9
15:09 jnettlet 3.10.9
15:10 jnettlet hste, I do not have an advancedsettings.xml Do you have a snippet of what should be there?
15:11 hste jnettlet: http://pastebin.com/Gg5bdmD8
15:14 jnettlet think I found my playback problem. ERROR: VpuOpen - VPU load failed with error code 1.
15:17 jnettlet hste, okay with dirtyregions patch I am running at about the same cpu usage and getting about 90fps
15:17 jnettlet really it is bouncing between 80-90 but mostly hanging out around 89-90
15:18 hste thats not bad for such a tiny device
15:18 jnettlet teh cpu seems to be bouncing wildly. I wonder if it is scaling too aggressively
15:19 jnettlet ha, yeah it is running efficiently enough that it is scaling the cpu back to the lowest frequency and then ramping it back up.
15:20 hste jnettlet: what about temp?
15:21 dv what are you playing?
15:21 jnettlet dv, just the gui
15:21 jnettlet the decoding is broken :-)
15:21 jnettlet well it can't open the device. I have to investigate
15:21 hste jnettlet: remember the firmware?
15:22 jnettlet okay so removing the lowest frequency form the scaling range has the device mostly chugging at 35% at 800Mhz
15:22 dv jnettlet: did you install the firmware?
15:22 dv you need one for the VPU
15:23 jnettlet I think I have it installed. But it probably wasn't loaded because I have the driver built into the kernel, but not the firmware.
15:25 jnettlet hste, temp is reporting 67C
15:26 jnettlet surface of the chip is reading about 34C
15:26 hste jnettlet: Do u have "VPU initialized" in dmesg?
15:26 jnettlet mxc_vpu 2040000.vpu: VPU initialized
15:27 jnettlet /dev/mxc_vpu is there and permissions are correct
15:27 dv jnettlet: to double check that everything is okay you could also try out the 0.10 gstreamer VPU plugins. or mine.
15:28 dv to rule out that it is a issue with the VPU in general, as opposed to a xbmc specific thing
15:28 hste jnettlet: do u have the udev rules http://stende.no-ip.info/files/10-imx.rules
15:28 jnettlet hste, no but I set the permissions manually for now
15:29 hste ok
15:30 hste jnettlet: U could try my ubuntu 13.10. vpu is working there
15:31 jnettlet ummm. The can the vpu do xvid decoding? I was going to through something at it that would be done with software.
15:31 jnettlet s/through/throw/
15:33 hste jnettlet: you could try big_buck_bunny_1080p_h264.mov
15:33 hste http://www.bigbuckbunny.org/index.php/download/
15:33 jnettlet I can try. But AAC is a bear to decode in software
15:35 dv the best aac decoder I have seen so far is the one from libav
15:36 dv much much faster than faad
15:36 dv (and available in the LGPL build of libav)
15:38 jnettlet ah something is broken. Still trying to send that through the VPU.
15:38 dv AAC through the VPU?
15:38 jnettlet must be some way to just use software decoding.
15:39 jnettlet oh I bet the .mov's video is h264
15:39 jnettlet dv, the vpu can't do AVI can it?
15:39 dv the VPU can do mpeg4p2, h264, h263, vp8, vc1 . it presumably can also do xvid and divx, which is a little unclear to me, because xvid/divx are just mpeg4p2..
15:39 dv the VPU cannot demux
15:40 dv ah, and the VPU can do mpeg1 and mpeg2 too
15:41 dv for h264, the VPU decoder expects the h264 bitstream use the byte-stream prefixes before the NALUs
15:41 dv plus, it also expects access units
15:45 jnettlet oh it can't find the vpu firmware file.
15:47 jnettlet this should be using /lib/firmware/vpu/vpu_fw_imx6d.bin right?
15:47 dv one thing that I am a bit concerned about with the xbmc port is the frame reordering in h264. the reordering can go back up to 17 frames iirc, so you'll need at least that number of allocated frames, plus you have to be able to deal with reordered frames at all
15:47 dv hmm lemme check
15:48 dv /lib/firmware/vpu/vpu_fw_imx6d.bin
15:48 dv yes
15:54 jnettlet haha awesome. I turned on vpu debugging it still just tells me that it can't find the file in folder...blah blah blah. Don't bother tell me what you are trying to open where
15:56 hste maybe it want the vpu_fw_imx6q.bin also :)
15:58 jnettlet nope it wants vpu_fw_0.bin. so mxc_cpu() is returning nothing useful
15:58 hste jnettlet : mine is -rw-r--r-- 1 root root 253968 Dec 11 13:44 vpu_fw_imx6d.bin
15:59 jnettlet that is getting hardcoded for now.
16:00 jnettlet oh that really made it unhappy
16:01 hste what happened?
16:07 jnettlet VPU firmware platform isn't matched
16:09 jnettlet okay imx guys any ideas?
16:11 hste jnettlet: you can check with my vpufw http://stende.no-ip.info/files/vpufw.tgz if its sth wrong with yours
16:13 jnettlet okay. there was another place I needed to short circuit the firmware for now. Got it to think it was decoding but still no video
16:14 jnettlet and no VPU_CODE_IRQ interrupts registered
16:15 jnettlet we have the same firmware
16:16 jnettlet okay well that is good progress. I need to feed the dogs and then screw around with gstreamer to see if I can debug this a bit more.
17:41 jnettlet at 720p the C1 can do 150fps with the xbmc gui
17:41 jnettlet vpu decoding is definitely broken. I am getting an error 7 no data available
19:11 Matoking Any information about possible camera support in Myriad Alien (aka Android support)?
19:11 Matoking Oops, wrong channel
22:10 pepedog Rabeeh?
22:23 pepedog Not in front of gear at the mo, where do I get build 6622 of xf86-video-imx-xivante? Got same version number as kernel but build number different
22:23 pepedog Segfaults
22:24 pepedog Builds and runs ok on utilite
22:56 hste jnettlet: solved the vpu problem?