IRC log of #cubox of Wed 12 Jul 2017. All times are in CEST < Back to index

15:48 topi`_> should "uEnv.txt" on the first vfat formatted partition of the sdcard be enough to instruct uboot correctly?
15:48 topi`_> first I got this from uboot: "** Unable to read file /boot/uEnv.txt"
15:49 vpeter> Put it in boot folder?
15:49 topi`_> then I created file boot/uEnv.txt on the first partition and retried. Now that mesage is gone (it says 60 bytes read in 11 ms"
15:49 topi`_> Importing environment from mmc0
15:49 topi`_> but still it tries to do netboot
15:49 jnettlet[m]> it checks boot/uEnv.txt first, then tries /uEnv.txt
15:50 jnettlet[m]> u-boot is just stupidly verbose about this
15:50 topi`_> instaed of mmcboot, which I wanted (I put something like "mmcboot=setenv bootargs blaablaa" into uEnv.txt
15:50 topi`_> I cannot check the current environment vars because uboot doesn't recognise my Microsoft Wired 600 keyboard
15:51 topi`_> I'll set up a serial console
17:12 topi`_> ## Executing script at 17f00000
17:12 topi`_> Wrong image format for "source" command
17:12 topi`_> what does this message mean? and where does this "script" at 17f00000 come from?
17:19 vpeter> topi`_: fatload loads it at that address. If we are talking about uEnv.txt.
17:27 topi`_> right
17:27 topi`_> but why is it wrong image format?
17:27 topi`_> uenv.txt is just a text file.
18:16 topi`_> jnettlet[m]: can one simply disable core(s) from imx6 if/when the environment gets too hot? a heatsink is very limited if the ambient temps around it get very hot
18:16 topi`_> I think disabling cores would be the first method of bringing down the chip temps
18:17 topi`_> it will of course affect somewhat the perf of our display, but it would still convey critical information
18:17 jnettlet[m]> topi`_: is your board not scaling the temps?
18:17 topi`_> should be
18:17 topi`_> I shall test :)
18:17 topi`_> but I think dropping cores would be even more effective
18:18 topi`_> in the summer the devices might face some very hot temps
18:18 jnettlet[m]> with our kernel both rabeeh and myself have had all the cores full out with neon instructions and using a heat gun and weren't able to overload it
18:18 topi`_> :)
18:18 topi`_> I think the real test is putting the device in a heating chamber and adjusting ambient temps to 60 degrees
18:18 jnettlet[m]> I spent a while on the cpu scaling code to be able to handle pretty much worst case scenario
18:19 topi`_> I guess I don't get that scaling code in the mainline kernel?
18:19 jnettlet[m]> we can test that if you would like
18:19 topi`_> we wanted the mainline for the etnaviv stuff
18:19 jnettlet[m]> no, I am trying to sort that out.
18:19 topi`_> not just etnaviv, but etnaviv and the newfangled DRM/GBM stuff
18:19 jnettlet[m]> mainline is the new focus.
18:20 jnettlet[m]> yeah, so with mainline you probably aren't getting that. I had some patches around at some point. I will dig them back up.
18:20 topi`_> etnaviv is somewhat buggy but nothing that we couldn't workaround
18:21 topi`_> btw I just tested openssl 1.1.0 from debian/stretch on a ThunderX vps
18:21 topi`_> aes128 speeds were 450MB/s
18:21 topi`_> I cannot wait for the imx8 stuff to arrive...
18:22 topi`_> the ThunderX doesn't have much bang in single core... maybe somewhat better than cortex-53 but not much. But that AES speed...
18:22 jnettlet[m]> the 8040 should be able to beat that then. The A72's should be much faster than the ThunderX cores
18:22 topi`_> problem with scp is that it usually limits transfer speeds at around 10-20MB/s because aes is so slow
18:23 topi`_> probably, gonna test that as well but right now I needed to borrow the power supply from macchiato over to the hummingboard
18:24 jnettlet[m]> you should also test rsync over ssh. that tends to get much better throughput
18:25 topi`_> yes if transferring lots of small files
18:26 topi`_> grr, the kernel's still compiling
18:26 topi`_> I need to move the compilation of HB kernels over to the macchiatobin :)
18:28 jnettlet[m]> in my notes on fast file transfer. I have a project called fast scp, which changes some buffers and such. I also have a link to this project which I never got around to testing because it uses rust. https://github.com/mcginty/shoop
18:28 jnettlet[m]> I have notes that it is interesting but infant
18:29 jnettlet[m]> fast scp was annoying because you need a special setup on the host machine.
18:31 topi`_> rust should be easy to build :) but it can get tricky if the project is using old libs etc
18:31 topi`_> they've moved stuff around in the language and as a result something that used to be in the stdlib can now be in an external lib and so on
18:31 jnettlet[m]> well if you give it a go, I will be interested if the numbers are that much better.
18:33 topi`_> Rust is interesting because it gives you the tools to write perfect threaded code
18:36 topi`_> btw I have an ATX power supply from a PC. But it seems it does not start feeding power until the user presses the "power" button on the PC
18:36 topi`_> how can I emulate that signaling?
18:37 topi`_> I think that signal is in the ATX poewr connector
18:37 jnettlet[m]> yeah, you jumper 2 of the pins
18:37 topi`_> seems the ATX power conn on the macchiato is the 24-pin version
18:37 jnettlet[m]> http://www.instructables.com/id/How-to-power-up-an-ATX-Power-Supply-without-a-PC/
18:38 topi`_> is it the PWR_OK or PS_ON# ?
18:39 topi`_> oh, the PS_ON# is that green one
18:39 jnettlet[m]> yes PS-ON
18:40 topi`_> so there is nothing on the macchiato motherboard that can be used to trigger this?
18:41 jnettlet[m]> no, I believe the only things wired up to the header is the 12V
18:41 topi`_> right
18:42 topi`_> I'll kludge that with a thin wire
18:42 jnettlet[m]> the pins on the bottom of the case are very accessible, just remember to tape over them
18:43 jnettlet[m]> sorry bottom of the board.
18:43 jnettlet[m]> answering two different questions :)
18:43 topi`_> :)
18:43 topi`_> what was the make target for linux kernel that created .deb packaging?
18:43 topi`_> make bindeb or something
18:43 jnettlet[m]> I think Artox answered that.
18:44 topi`_> yeah
18:46 topi`_> make bindeb-pkg
18:46 topi`_> my memory is getting so bad that I should retire soon...
18:47 jnettlet[m]> or take better notes. Without my personal kb I would be doomed
18:47 topi`_> yeah I need a big notebook :)