IRC log of #cubox of Wed 04 Dec 2013. All times are in CET < Back to index

00:59 _rmk_ lioka: here?
00:59 _rmk_ lioka: your compat patch - care to add a signed-off-by to it please?
09:33 lioka _rmk_: sure. http://ns.lioka.obninsk.ru/0001-compat-api-stuff-added.patch
11:59 _rmk_ lioka: thanks
12:01 Bluerise _rmk_: Hey. :) Any news on the BSD-license FDT stuff?
13:15 SYS64738 hi
13:21 MarcusVinter Hey
13:28 SYS64738 ha anyone tried this: http://wiki.xbmc.org/index.php?title=HOW-TO:Compile_XBMC_for_Linux (ubuntu 12.04 LTS) on cubox ?
13:28 SYS64738 has
13:59 MarcusVinter Not done it myself but it looks pretty straight-forward.
14:00 MarcusVinter Why? Are you stuck on something?
14:56 jnettlet SYS64738, if you do the geexbox install via the cubox installer that has a customized xbmc build included.
15:16 erikohman Hi, i'm just wondering the shipping capacity per day? do anyone know this?
15:25 jnettlet hi, I know that nobody knows me, but could you tell me all the details about your supply chain and production capacity? :-P
15:32 rabeeh :)
15:32 rabeeh jnettlet: erikohman?
15:33 jnettlet rabeeh, yeah. just flipped over to the channel and saw that.
15:34 jnettlet Of course it is good because people are excited to get their product.
15:37 rabeeh yeah. people are working like crazy here to get the systems out of the door ASAP
15:45 dv_ \o/
16:25 neofob so `openssl speed -evp aes128` gives me -infk (I can't recall exact string)
16:26 neofob `openssl speed -evp aes128 -elapsed` gives me about 13MB/s on 8K blocks
17:26 hste dv_: Did u see this https://community.freescale.com/thread/314986
17:46 dv_ hste: thanks
17:47 hste dv_: or maybe u r their development team :)
17:47 dv_ :D
17:47 dv_ well it wouldnt surprise me if they have a team somewhere quietly working on this
17:48 dv_ freescale seems to have management problems where the left hand does not know what the right one is doing
17:49 rabeeh dv_: i agree
17:49 rabeeh it's so fragmented work
17:53 hste and not any focus on using the soc in a desktop environment
17:54 hste but at least they have a community site where question can be asked :)
17:56 rabeeh dv_: added comment
17:56 rabeeh hste: this is great; but not only support questions; they need to embrace the community
17:57 dv_ rabeeh: thanks for the endorsement :)
18:02 neofob rabeeh: so why is the number of openssl with and without -elapsed parameter so different on cubox?
18:02 neofob openssl speed test, i mean
18:04 rabeeh bug?
18:08 rabeeh neofob: i can't find elapsed i nthe openssl man page
18:10 neofob rabeeh: but openssl accepts that parameter
18:10 neofob it says something about using cpu time instead of timer
18:11 neofob i do see speedup with cryptodev but running openssl w/o -elapsed parameter can give you a false hope :D
18:11 neofob in the end, i don't know which number is more reliable :/
18:14 jnettlet neofob, it is not what it uses, but what it reports. with the -elapsed option the speed test reports the actual amount of time the test took to run. Wihtout it it reports how much cpu time was spent running the tests.
18:14 neofob arh, sorry i got it backward; -elapsed measures the elapsed time instead of user cpu time
18:14 neofob jnettlet: thanks
18:15 jnettlet running without -elapsed should give you a much lower result because it should be using the crypto engine instead of the cpu
18:15 jnettlet I guess both should give you lower results because hopefully the test runs faster, but -elapsed should be longer than without it
18:54 erikohman Hi, im just wondering what the shiping capacity per day is atm?
19:05 jnettlet hmmm. I wonder how many times he is going to show up ask that question and then leave 5 minutes later between now and Monday.
19:22 hste jnettlet: I guess maersk have big shipping capasity :)
19:41 _rmk_ jnettlet: yes, quite silly behaviour.
19:44 hste jnettlet: looks likes its getting windy in denmark and norway tomorrow
19:45 _rmk_ hste: and Scotland
19:46 hste _rmk_: yes I guess it will be windy there too :)
20:52 _rmk 20:52 * _rmk_ pulls lots of code around in xf86-video-armada... now pulled most of the nonspecific DRM code into a separate file
20:53 _rmk_ armada_drm.c is now 640 lines. Who needs more? :)
20:56 jnettlet hste, is it? I generally don't look at the weather until the morning.
20:56 _rmk_ yea, over here they're saying its our first proper winter storm, winds gusting to 75-80mph
20:57 jnettlet _rmk_, oh I have your vivante acceleration code pulled into int's own Xorg extension. I am just playing around with integrating that with the vivante GL acceleration DRI implemetation.
20:58 _rmk_ have you come to any conclusion about the v4 drivers via freescale?
20:59 jnettlet I think we can use them. I will probably put a disclaimer on the download site or included that says the drivers are only for use on Vivante hardware, and can not legally be decompiled blah blah blah blah blah blah
21:01 jnettlet But Freescale's appendix A license, which I missed in my initial hasteful reading, seems to mesh up with the Marvell license of the binaries.
21:01 jnettlet Free to use on Vivante hardware and redistribute without modification.
21:02 jnettlet Not sure what other hardware you would use them on :-)
21:02 _rmk_ an FPGA? :)
21:03 _rmk_ okay, that's good, it means I can think about updating xf86-video-armada to use the v4 stuff
21:03 _rmk_ or your stuff
21:04 jnettlet well the good news is I can take a step back and rip out the compatibility layers I have been f'ing around with and just focus on a single "stable" api.
21:05 _rmk_ rotfl "stable" :)
21:05 jnettlet Then it is just a balancing act of handling 2-3 dri drivers in userspace.
21:06 _rmk_ the alternative is I put the knowledge I have of the 2d vivante stuff to good use and push etnaviv forward
21:08 jnettlet Our driver model really needs to be. KMS/FB frontend allocates scan out buffers and some small buffers maybe for cursors and glyphs. Then we pass those into the GPU driver which provides the acceleration architecture and can allocate it's own pixbufs. I have most that sorted out.
21:10 jnettlet The v4 binary vivante driver has this added layer of supporting GL on it's GLESv2 hardware which is intriguing for desktop usage. Especially since most distros are still not properly compiling 3d libs with the proper backend support.
21:11 jnettlet in theory I can designate the vivante dri driver to handle all 3d windows and then have everything else accelerated through my vivaccel extension based on your 2d work.
21:12 jnettlet then armada/imx/etc can use whatever framebuffer kms driver they want as long as it can pass some dmabuf fd's into the vivaccel extension.
21:13 jnettlet This was always so much easier when you just had videoram and system memory, a PCI/AGP bus and a GPU
21:20 dv 21:20 * dv_ is confused
21:20 dv_ what is this about the v4 drivers?
21:44 jnettlet dv_, that is the latest vivante binary driver version.
21:45 dv_ ah
21:45 dv_ so your architecture would allow for pluggin in these, and in future some replacement, like robclark's drm
21:46 dv_ I mean, support for it (and etnaviv)
21:54 jnettlet that is what I am hoping.