| 02:54 | heap_ | so |
| 02:54 | heap_ | i have a esata storage and im copying from lvm to physical drive ... and its 5mb/s ??? |
| 02:54 | heap_ | 5mb/s???? |
| 02:54 | heap_ | is that real or dream? |
| 02:57 | heap_ | iam really dissapeared |
| 02:57 | heap_ | from 2 drives lvm to another separate drive... why its 5mb/s? |
| 02:57 | heap_ | all 3 drives are inside one box (->esata to cubox) |
| 02:58 | heap_ | rsync -avz --progress -e 'ssh -c arcfour128' --exclude /mnt/esata/usb-4TB/ /mnt/esata/ . |
| 02:58 | heap_ | 1,318,158,336 16% 4.70MB/s 0:22:39 |
| 02:58 | heap_ | ;-((( |
| 03:05 | cbxbiker61 | it may not be obvious to you heap_... but you're driving the data across one e-sata connection, so yeah you shouldn't expect great data rates |
| 08:49 | jnettlet | heap_, okay so I did a quick test on my profiling box. The bottleneck you are seeing is illustrated here. http://pastebin.com/sK73VjCk |
| 08:49 | jnettlet | the biggest bottleneck is not the hardware or encryption, it is actually the compression. |
| 08:51 | jnettlet | by default rsync uses its own internal zlib version for compression. I am recompiling my debian version to use an external zlib library which I have patched to include neon acceleration for the adler32 functions. I will see how that speeds things up. |
| 08:52 | jnettlet | the other problem with this method is that rsync is single threaded so you are bound by the performance of a single core. |
| 08:54 | jnettlet | actually just removing the compression and still using ssh with caam crypto acceleration I am seeing performance around 25MB/s |
| 08:55 | jnettlet | that is using aes-256 |
| 09:58 | heap_ | jnettlet http://pastie.org/10554269 |
| 10:00 | heap_ | jnettlet: oh yes, i removed -z option and its goes 25mb/s |
| 10:00 | heap_ | even 35mbs/s |
| 10:00 | jnettlet | yep |
| 10:00 | heap_ | jnettlet: thanks |
| 10:00 | heap_ | jnettlet: any idea about that pastie? |
| 10:01 | jnettlet | this is running debian with my kernel? |
| 10:06 | heap_ | jnettlet: arch arm linux < |
| 10:06 | heap_ | ill install that debian today |
| 10:07 | jnettlet | yeah can't help you then. They do all sorts of odd things to their kernels. Quite a few arch users have been sent here with problems that we can't do anything about. |
| 10:07 | heap_ | oh ok |
| 10:08 | heap_ | ill try your distro + kernel |
| 13:52 | topi` | I've got a HB Gate here that won't show anything on the HDMI. I need to hook up with its serial console - anyone know what PINs and in which conector? |
| 13:52 | topi` | the HB Classic was easy, since the GPIO26 layout was the same as with the raspberrypi |
| 13:52 | topi` | GND to pin 5, and then TX/RX were in pins 7, 9 |
| 13:54 | topi` | I'm studying the schematics, and I can only see some camera/display pins in the 36 pin GPIO header... |
| 13:55 | topi` | but it says cryptically that "by default the linux kernel device tree sets those pins as GPIOs" |
| 13:55 | topi` | ooh, J25 is labeled as "main debug console" |
| 13:56 | topi` | now I just need to locate it on the board :) |
| 13:57 | topi` | found! |
| 13:58 | topi` | is this 3.3V uart? |
| 13:58 | topi` | I assume so, because that uart circuitry is surrounded by VCC_3V2 signals |
| 14:02 | topi` | ah, the Gate won't boot because of error -110 transferring data (from MMC) |
| 14:07 | jnettlet | topi`, if the SDHC card is on the upper left corner then the console uart is the first set of 3 pins to the right with ground on the left |
| 14:08 | jnettlet | topi`, which sdhc card? |
| 14:08 | jnettlet | and which distro/image |
| 14:10 | topi` | I grabbed an image named "sr-imx6-debian-jessie-cli-20151019.img.xz" |
| 14:10 | topi` | as suggested by Artox |
| 14:10 | topi` | but that SD card gave me the same error -110 from the same sector # when I inserted it into my Chromebook... |
| 14:10 | topi` | so I threw the bloody card to trash |
| 14:10 | topi` | next :) |
| 14:11 | topi` | that was Transcend brand, maybe I need to use only Sandisks |
| 14:13 | topi` | snd_soc_register_card failed (-5) ... is this because the Gate doesn't have audio soldered in? |
| 14:13 | jnettlet | topi`, was it a UHS-I card? |
| 14:13 | topi` | no |
| 14:13 | topi` | some old 2 GB card |
| 14:13 | jnettlet | topi`, oh..so probably done for. okay |
| 14:14 | topi` | heh |
| 14:14 | jnettlet | yes that error is because of the lack of audio codec. I am working on a new u-boot that will properly patch device-tree for the different boards. |
| 14:14 | topi` | OK |
| 14:15 | jnettlet | fyi, that is how the HB works as well if you have the base vs the pro. |
| 14:15 | topi` | odd, I see the UBOOT stuff in the minicom, but cannot interrupt it so I presume it doesn't see my characters sent |
| 14:16 | jnettlet | topi`, turn off hw flow control in minicom |
| 14:16 | jnettlet | or whatever terminal program you are using |
| 14:16 | topi` | it is off |
| 14:16 | jnettlet | that is odd, that is usually the behaviour |
| 14:16 | topi` | maybe I'll try using screen with the serialport |
| 14:17 | topi` | same thing :( |
| 14:17 | topi` | damn, wrong baudrate with screen |
| 14:17 | jnettlet | you have ground connected to the pin with the triangle? |
| 14:18 | topi` | yeah |
| 14:18 | topi` | without it, wouldn't receive the TX's :) |
| 14:19 | topi` | umm, the triangle is pointing to pin 3 and not pin 1! |
| 14:19 | topi` | in the schematics, GND is in pin 1 |
| 14:20 | topi` | need to handle kids 1 and 2, bbl |
| 14:20 | jnettlet | triangle is ground, so that explains the no TX |
| 14:27 | topi` | OK i'll try again :) |
| 14:32 | topi` | it works! |
| 14:32 | topi` | uh, only 250MB left of RAM (512M) after booting the cli version of jessie... how much does the video buffers take RAM? |
| 14:33 | topi` | on the Raspberry, video eats up 64M of ram |
| 15:11 | topi` | I need to compile my custom kernel in order to get the stupid logitech keyboard to work (it's a kind of "combo" device where one controller connects to both kbd and mouse) |
| 15:11 | topi` | unfortunately the microsoft usb keyboard is not here... |
| 15:13 | topi` | the image seems to only have 1 partition... where's the "boot" partition? |
| 15:13 | jnettlet | it does have only one partition, that is just how debian does it. |
| 15:13 | topi` | ok, so where are the UBOOT and the kernel? |
| 15:13 | topi` | and uEnv.txt? |
| 15:14 | jnettlet | if you are missing kernel modules please file bug reports on github.com/SolidRun/linux-fslc |
| 15:14 | jnettlet | in /boot :) |
| 15:14 | topi` | damn how I hate the fact that I cannot mount ext4 on OSX:.. |
| 15:15 | topi` | I need to get my Chromebook |
| 15:16 | topi` | OK, so UBOOT is able to mount that root partition and look at /boot inside? |
| 15:17 | jnettlet | yep |
| 15:17 | topi` | fair enough. |
| 15:25 | topi` | hmm, libvdpau.deb was upgraded, but does this mean that there is working VDPAU acceleration for imx6? |
| 15:25 | jnettlet | not at all |
| 15:25 | jnettlet | that get pulled in by gstreamer I believe |
| 15:25 | topi` | OK |
| 15:26 | topi` | so I just use gstreamer in conjunction with the imxipusink_ |
| 15:26 | topi` | ? |
| 15:26 | jnettlet | vdpau and vaapi won't work with the existing hardware because it does bitstream decoding rather than by slices |
| 15:26 | jnettlet | you can use autovideosink, the choices are to use the ipu, g2d, or egl, all three should be vsync'd now |
| 15:27 | topi` | ok |
| 15:34 | topi` | I'm cloning from github.com/solidrun/linux-fslc, I presume this is where the kernel is baked from? |
| 15:34 | topi` | when building those sr-* images |
| 15:34 | jnettlet | yeps |
| 15:35 | topi` | my build machine is a Tegra K1 devboard :) |
| 15:35 | topi` | the Tegra X1 board is out now, but I cannot afford $599 |
| 15:35 | topi` | not worth it :) |
| 15:36 | jnettlet | get a nvidia shield then :) |
| 15:36 | topi` | which defconfig for the HB Gate /solo ? |
| 15:36 | jnettlet | although I don't dare to hack mine yet |
| 15:36 | jnettlet | same as all the other. |
| 15:36 | topi` | OK |
| 15:36 | jnettlet | it is a hb after all |
| 15:37 | topi` | what's the difference between v7_cbi_hb_base_defconfig and cbi_hb_defconfig? |
| 15:37 | topi` | not related to the existence of HB Base, I presume... |
| 15:41 | Artox | topi just the feature set |
| 15:42 | jnettlet | base has just core device support |
| 15:42 | Artox | I always build the nonbase defconfig for my packages |
| 15:42 | jnettlet | for the board |
| 15:42 | jnettlet | hey Artox, do you know you have fallen of oftc.net? |
| 15:43 | Artox | jnettlet: Greetings. Nope didnt notice. Will reconnect some time later tonight. |
| 15:43 | jnettlet | okay |
| 16:40 | Tenkawa | is the cubox-4i any good for transcoding dvds |
| 16:40 | Tenkawa | ? |
| 16:43 | jnettlet | Tenkawa, good question. It would have no problems doing the encoding part, I am not sure how fast it could crack the encryption. |
| 16:45 | Tenkawa | ahh |
| 16:45 | Tenkawa | yeah arm in general has a disadvantage |
| 16:45 | Artox | oh? |
| 16:46 | Tenkawa | according to what i've read anyways |
| 16:46 | Tenkawa | I just want a decent all purpose server anyway |
| 16:47 | Artox | oh that famous thing |
| 16:47 | Tenkawa | and the 4i pro looks like a really good debian candidate for me |
| 16:47 | Artox | candidate for sure |
| 16:47 | Tenkawa | can it run the 4.x series kernels? |
| 16:47 | Artox | ye |
| 16:48 | Tenkawa | cool |
| 16:48 | Artox | but |
| 16:48 | Tenkawa | might go pick up one today (actually got a local vendor) |
| 16:48 | Artox | video en-, decoding and other nice things are not fully mainlined |
| 16:48 | Tenkawa | true |
| 16:48 | Tenkawa | thats ok |
| 16:48 | Tenkawa | thats just a nice to have for me |
| 16:49 | Tenkawa | I'm really looking for something to be my server that has more multipurpose io than my current one (pi2) |
| 16:49 | Tenkawa | I'd really like to hook up my ssd sata, etc |
| 16:51 | jnettlet | Tenkawa, where is your local vendor? |
| 16:51 | Tenkawa | ever heard of microcenter? |
| 16:52 | jnettlet | lol... So you are in Cambridge, Chicago, or SanJose |
| 16:52 | Tenkawa | usa (if you arent in usa) |
| 16:52 | Tenkawa | jnettlet: nope |
| 16:52 | jnettlet | they expande? |
| 16:52 | Tenkawa | yeah alll over |
| 16:53 | jnettlet | wow, it has been a while since I was in the US. I used to live not far from the first Microcenter in Cambridge, MA. Opened while I was back in university |
| 16:53 | topi` | damn it takes a LONG time to compile the non-base defconfig with make -j6 |
| 16:53 | Tenkawa | about 25 stores now |
| 16:54 | topi` | there's a huuge amount of modules |
| 16:54 | Tenkawa | jnettlet: its become "the" place to get non-traditional computer/electronic stuff |
| 16:54 | Tenkawa | topi`: yep |
| 16:56 | topi` | jnettlet: did you find out if freescale is already shipping the "improved" i.mx6, i.e. i.mx6+ ? |
| 16:56 | topi` | with better bandwidth, etc |
| 16:58 | jnettlet | topi`, haven't seen them |
| 16:58 | jnettlet | they were supposed to be released in August |
| 16:59 | jnettlet | so they are right on schedule for Freescale |
| 17:06 | Tenkawa | jnettlet: think the 4i-pro will be the main model for a while still |
| 17:06 | jnettlet | Tenkawa, I wish I knew. I was hoping to have the new chips out already |
| 17:13 | topi` | jnettlet: how much do you think is the latency from freescale chip production/shipping to the manufacturing of SR SOMs? |
| 17:13 | topi` | maybe current SOMs are being built from a stockpile of i.mx6 parts |
| 17:14 | jnettlet | the iMX6 is an embedded grade chip which means it will be fabricated for 7 years. We are about half way through the support lifetime |
| 17:14 | topi` | yeah that's a long-term part |
| 17:15 | topi` | on the Raspberry, video eats up 64M of ram |
| 17:16 | topi` | uh, only 250MB left of RAM (512M) after booting the cli version of jessie... how much does the video buffers take RAM? |
| 17:16 | topi` | (I want to re-ask this :) |
| 17:17 | Artox | topi`: As far as I know the CMA allocates 256M by default |
| 17:17 | Artox | but it will give the memory to other processes if it is still unused when they need it |
| 17:18 | topi` | OK |
| 17:18 | topi` | I was just looking at "free's" output and it seemed that 250M was already used |
| 17:18 | Artox | your best test would be writing a program that triesd to allocate memory |
| 17:19 | Artox | and see how much you get |
| 17:19 | topi` | good idea |
| 17:19 | jnettlet | the gpu will grab 128M at boot |
| 17:19 | jnettlet | but you should also see what your cached/buffered memory size is |
| 17:19 | topi` | what if we don't use the GPU, only the imxipusink? |
| 17:20 | jnettlet | then you best bet is to disable it in device-tree |
| 17:20 | Tenkawa | jnettlet: yeah just trying to determine if i should wait for a new gen or if its likely to be a while |
| 17:21 | jnettlet | Tenkawa, sure. If you are using it for server type things then your won't get too much additional benefit |
| 17:21 | jnettlet | you do have the bonus that you can upgrade just the microsom with our hardware. |
| 17:22 | Tenkawa | cool |
| 17:22 | topi` | that's quite clever, since the CPU/GPU/RAM is there |
| 17:22 | topi` | and baseboard components are, well, more time-tested |
| 17:28 | Tenkawa | bbl.. cheers all |
| 19:07 | jnettlet | well if Tenkawa comes back he can be happy to know that DVD ripping shouldn't be a problem. I just enabled DVD playback from my usb dvd rw and it seems to be working just fine |
| 19:08 | jnettlet | the playback takes about 65% of one core, so on a dual-core there should be no problem transcoding it in real time |
| 19:13 | Exaga | http://solid-run.com/marvell-armada-family/clearfog/ = oh hell YES! |
| 19:13 | Exag | 19:13 * Exaga grins like a chesire cat |
| 19:15 | cbxbiker61 | hey jnettlet, can you point me at the kodi patches for the new api? |